Strong claims shouldn’t hinge on a single study, but this study can be used to update probabilities for certain theories. You can think of it as eliminating possibilities.
That’s my issue — I don’t think Padian et al.’s zero seroconversions result is sufficiently strong to outright eliminate a possibility, unless one’s already assigned a low prior to that possibility. Personally, I walked into this discussion with a very high prior (something like 99%) for HIV being sexually transmissible to some nontrivial (more than, say, once in every million unsafe sex acts) extent, and the Padian et al. result is not a strong one, so my prior is essentially unchanged.
Namely it eliminates or vastly reduces any hypotheses involving a typical STD transmission route. It allows for a very small possible horizontal transmission rate, but the most likely cluster is centered around zero.
Since the prior estimates for sexual transmission I’ve seen are already very small (on the order of 0.01% to 0.1%), the Padian et al. result doesn’t really clash with them that much. I also doubt “the most likely cluster” is centred on zero if one accounts for all of the evidence in the study, rather than focusing exclusively on the zero seroconversions result.
That doesn’t prove it is zero, it is just that this evidence strongly favors zero transmission rate theories over all others.
It favours a zero transmission rate strongly only if one makes particular strong assumptions about the sexual behaviour of the study’s subjects, and disregards the evidence of risk factors.
The what? I don’t think there is a single “female HIV seroconversion” to be “explained away” in any of those studies.
Let’s return to one of the papers that kicked off this discussion, as it has a few examples. From page 810: “seroconversions occurred in three male and seven female contacts after enrolment in the study” (that’s not even counting whatever seroconversions had to have occurred in those subjects already HIV+ at the start of the study). There must be some explanation for those seven seroconversions.
From what I understand, and what Padian et al claim, the Padian study is the only properly controlled prospective study of heterosexual transmission. The other studies are just observational guesswork that are all circularly dependent on the notion that seropositivity is transmissible—they see a couple where both test positive and they just assume it was sexually transmissible—hardly anything resembling evidence for the theory—just circular logic.
Whether or not one believes HIV is sexually transmitted, seroconversions occurred. What were their causes, if not HIV infection due to sexual transmission?
First of all, in the other theories, there is no transmission to explain—and it certainly doesn’t have to involve iv drug use.
Where does the HIV come from if not an external source?
I glanced at the Madrid study and find it largely worthless compared to the Padin study. It is small, not properly controlled, and not prospective.
I do not see the Madrid study as “largely worthless”, though I’d give it less weight than the Padian et al. study for the reasons you give. (I should emphasize that I’m talking about the Padian et al. study as a whole there; I maintain my reservations about the zero seroconversions result.) Still, the Madrid study sample’s not particularly small compared to the number of Padian et al. couples who consistently had unprotected sex and were followed up prospectively.
Your “ingenious study” is not well controlled for drug use, and in fact barely even mentions it.
Yes. The point I was making was that with the sequence analysis confirming the source of the transmission, the probability of any given seroconverter having been infected by IV drug use with strangers is less than if the researchers hadn’t done the sequence analysis (although of course still more than zero). So, to have me agree that the HIV infections are due to IV drug use, I would have to be convinced that the seroconverters had been sharing needles with their partners. And...
As HIV/AIDS is highly correlated with drug use in all the epidemological data I have seen in the west,
...as the study took place in Uganda, I find it very unlikely that even half of the seroconverters were infected by sharing needles (especially since none of the subjects reported injection drug use). Unless you are now arguing that seroconversion doesn’t actually indicate HIV infection (and I can’t tell whether you are or not), I expect you can see why I’d want alternative explanations for each of these seroconversions.
So to show that lung cancer is caused by a sexually transmitted vector, you would need to focus on non-smokers. That is why the Padian study is useful, and this African study you point to is not useful.
The African study, I’d say, is weaker evidence than the Padian study. At the same time, I think it’s absurd to write it off as “not useful”. At the risk of repeating myself, my reasoning goes like this: some number of subjects acquired HIV during the Ugandan study; exposure to HIV is necessary for infection; it is not credible to suppose that every HIV exposure that led to seroconversion is attributable to IV drug use. If there were only four or five seroconversions, I’d say it was statistically possible. In fact, the study reports that out of 239 monogamous couples that were initially HIV-discordant, 72 “acquired HIV during follow-up”. The probability that all 72 of these Ugandans acquired HIV by IV drug use is surely astronomically small.
If you want to get into the side discussion on “viral load” and viral DNA comparison based on PCR techniques,
Not especially, to be honest.
it’s critical that you understand Kary Mullis’s objection. Kary Mullis won the nobel for inventing PCR, and he is a strong critic of the whole HIV theory, and the entire interpretation of genetic data based on the technique he invented.
I’m not much impressed by Mullis’s Nobel credentials. There are quiteafewNobelists who got the prize and promptly started dabbling in crankery. (Including, as it happens, Luc Montagnier, so this isn’t just an issue for people who downplay the links between HIV & AIDS.) Jim Watson co-discovered DNA’s structure, but I’m not about to take all his proclamations about genetics at face value.
Anyway, I haven’t seen anything in the Uganda study report to suggest they used PCR in so sloppy a way as you’re suggesting, and I doubt they’d have much to report if they had. As such, it’s hard for me to avoid the feeling that your remarks about PCR are something of a fully general counterargument against HIV sequencing.
And finally, keep in mind that any way you slice it—we should expect to see sequence similarities in these couples, as they are genetically related—from the same area in africa and much more related on average than two random samples of homo sapiens.
Right, but two samples of virus from a couple where one member infected the other will nonetheless be far more genetically similar on average than two samples of virus from randomly selected infected people in Rakai.
Whether or not one believes HIV is sexually transmitted, seroconversions occurred. What were their causes, if not HIV infection due to sexual transmission?
First of all, in the other theories, there is no transmission to explain—and it certainly doesn’t have to involve iv drug use.
Where does the HIV come from if not an external source?
I’m having difficulty squaring some of your comments on seropositivity and HIV testing with the high reported sensitivity & specificity of properly conducted HIV tests.
First off, before we get into anything else, we need to understand and agree on what seropositivity actually means. Seropositivity means a positive result on an antibody test such as western blot, to some combination of antigens identified by Gallo in 1985 in his large scale blood screening tests. He did a large number of tests on transformed (immortalized) cell lines derived from AIDS patients and found a combination of antigens that could screen blood—separating AIDS and pre-AIDS blood from regular blood.
The problem is he (like Motagnier) failed to isolate the ‘virus’, and most or all of the antibodies in the test react or cross-react with antigens to opportunistic microbes (candida namely) and cellular debris. The p24 protein in particular is essentially just a normal cellular wall or microvescile component—so the ‘HIV’ test is really just a general test of opportunistic infection and apoptosis or immune directed killing of CD4 cells (possibly due to widespread viral parasite burden). It is not a measure of ‘HIV’, it is a direct measure of declining CD4 cells and AIDS or pre-AIDS.
The antibody tests are not standardized geographically or temporarily, so it also makes it very difficult to compare across studies—“seropositivity” means different things in different places and times.
As just one random example—most dogs typically have a mix of ‘HIV’ antigens, and are HIV ‘seropositive’ in whole or in part:
reported that 72⁄144 (50%) of dog blood samples “obtained from the Veterinary Medical Teaching Hospital, University of California, Davis” tested in commercial Western blot assays, “reacted with one or more HIV recombinant proteins [gp120--21.5%, gp41--23%, p31--22%, p24-- 43%]”
You post a link to a paper which supposedly shows the ” high reported sensitivity & specificity” of HIV tests. This is not actually what that paper is about, but it references several other papers for this claim—the first I investigated being this. The important quote:
Thirty-fourwomentested HIV-1 positive
with both rapid test and EIA, and all
were confirmed by Western blot
(prevalence=7/1000).
So they are just using Western blot as ‘confirmation’. So they are just using one antibody test to confirm another antibody test—which of course is rather ridiculous.
To actually compute the sensitivity and specificity for a “HIV” test, one needs a gold standard such as viral isolation or perhaps a DNA test. Unfortunately HIV can not be isolated, either because it doesn’t exit or it exists in only minute quantities.
But one can attempt to use the presumed viral DNA as a gold standard, and the result is extremely poor sensitivity and specificity:
Poor sensitivity is perhaps a gross understatement—the study actually shows that around 18-25% of the population at large test positive for ‘HIV DNA’, and this is only weakly correlated with seropositivity.
You completely dismiss Mullis’s argument based solely on an ad hominem “not much impressed by Mullis’s nobel credentials” without seeming to acknowledge or understand the argument itself.
Seroconversion in the west is closely correlated with AIDS or pre-AIDS. This does not appear to be as true in Africa, so we are generally talking about two different worlds. Part of this may be genetic (black americans have amazingly higher seropositvity in general), the other part may relate simply to higher precedence of opportunistic infections and antigens that seropositivity measures. KS for example is vary rare in the west and along with systemic candidaisis was part of the original AIDS definition, but it is one of the most common cancers in Africa.
Unless you are now arguing that seroconversion doesn’t actually indicate HIV infection (and I can’t tell whether you are or not),
I am arguing that.
Seropositivity does not strongly correlate with ‘HIV’ infection (by DNA test), which is why it is better to discuss AIDS itself as being sexually transmittable or not.
The Gallo blood test is tightly correlated with AIDS (at least in the west) - simply because that is what it was designed to do, so you can use that as data for AIDS transmission discussions.
Unless you are now arguing that seroconversion doesn’t actually indicate HIV infection (and I can’t tell whether you are or not),
I am arguing that.
OK. At this point, I’m going to have to disengage and walk away from this debate. I’m realizing that the inferential distance between us is far bigger than I originally thought, and trying to bridge it would need me to considerably ramp up the effort I’m already putting into this. (Even then I can imagine this going on indefinitely, which isn’t a very appealing prospect to me, nor to other Less Wrong posters, by the looks of it.)
I’d still like to respond briefly to one part of your comment, which comments on my own words rather than HIV/AIDS:
You completely dismiss Mullis’s argument based solely on an ad hominem “not much impressed by Mullis’s nobel credentials” without seeming to acknowledge or understand the argument itself.
It’s wholly legitimate for me to respond to someone citing Mullis’s credentials (as if I didn’t know about them already) by explaining why I give them little weight, and my next paragraph was meant to summarize why I gave “your remarks about PCR” (that is, those you paraphrased from Mullis) short shrift. In other words, I acknowledged the argument by rejecting it.
I’m glad you can walk away, I have a harder time initiating that. I’m curious though about the direction of the inferential distance you see—do you have a biology background?
The dissidents point to a rather surprising pile of evidence that the serological HIV tests are based on rather general, cross-reactive antibodies, and this is essentially a fundamental flaw in HIV science which has never been corrected. Now it may be that the orthodoxy has a really good counter to this, but if they do I have yet to find it. The orthodox position on this, from papers linked to wikipedia, points to studies which measure the sensitivity of various HIV antibody tests by comparing them to . . other HIV antibody tests.
The few large double-blind meta-studies that compare the different antibody tests to PCR tests show terrible sensitivity and specificity between the two, and I haven’t seen the orthodox counter to this. So something is wrong with the antibody tests, the PCR tests, or the whole thing. I imagine it’s a little bit of both—the antibody tests are cross-reactive (hence many dogs test positive), and PCR tests are difficult and subject to experimenter bias.
Perhaps the orthodox counter is that there are a whole big host of HIV related viruses, and the antibody tests are cross-reactive across these related species. This seems to then just beg more questions than it answers, and doesn’t circumvent some of the specific non-viral cross-reactions the dissidents point to.
My paraphrase of Mullis’s argument may actually be a mix of other dissident positions. I just rechecked that part of his book and he covers the difficulty of PCR and the confirmation bias but largely in regards to the OJ trial. On HIV he mainly rehashes Deusberg’s argument.
I’m curious though about the direction of the inferential distance you see—do you have a biology background?
None at all. (I expect the inferential distance would be even greater if I did. If I had personal experience of working with retroviruses, for instance, I reckon my prior probabilities for claims like “HIV can not be isolated” or “HIV doesn’t exist” would be far, far less than they are. And they are already very low.)
To actually compute the sensitivity and specificity for a “HIV” test, one needs a gold standard such as viral isolation or perhaps a DNA test. Unfortunately HIV can not be isolated, either because it doesn’t exit or it exists in only minute quantities.
When ‘HIV’ was first ‘discovered’ in the original papers by Gallo and Montagnier, they had difficulty isolating and didn’t publish pictures from what I understand—that didn’t happen until years later. Gallo’s great discovery for HTLVIII was based on running a lager number of antigen/antibody tests with an immortalized cell line to find an antibody test that could screen AIDS and pre-AIDS blood from regular blood. That is the basis of all the current HIV tests.
The first published pictures came more than a decade later, and they showed that “HIV isolate” really consists largely of cellular debris and microvesciles. In these EM photos, they do find some occasional particles of roughly the right size and label them as “HIV”, but they could also just be any of a number of other things, and for all intents and purposes, HIV ‘particles’ look like regular microvesciles.
The titles of the papers say it all:
“Cell membrane vesicles are a major contaminant of gradient-enriched human immunodeficiency virus type-1 preparations”
“Microvesicles are a source of contaminating cellular proteins found in purified HIV-1 preparations”
I have lost the link, but there are better more recent pictures taken with ATM, and they show that for all intents and purposes, it’s impossible to distinguish ‘HIV’ from regular microvesicles that bud from the cell wall naturally. If HIV can be said to exist at all as a unique exogenous virus, it is only because of unique RNA content in the microvesicle, and in this sense is very much unlike all other known viruses.
Of course, the part of HIV’s genome which is supposed to code for the outer envelope is pretty much the same as the endogenous sequences that already exist in the human genome—the HERVs.
That’s my issue — I don’t think Padian et al.’s zero seroconversions result is sufficiently strong to outright eliminate a possibility, unless one’s already assigned a low prior to that possibility. Personally, I walked into this discussion with a very high prior (something like 99%) for HIV being sexually transmissible to some nontrivial (more than, say, once in every million unsafe sex acts) extent, and the Padian et al. result is not a strong one, so my prior is essentially unchanged.
Since the prior estimates for sexual transmission I’ve seen are already very small (on the order of 0.01% to 0.1%), the Padian et al. result doesn’t really clash with them that much. I also doubt “the most likely cluster” is centred on zero if one accounts for all of the evidence in the study, rather than focusing exclusively on the zero seroconversions result.
It favours a zero transmission rate strongly only if one makes particular strong assumptions about the sexual behaviour of the study’s subjects, and disregards the evidence of risk factors.
Let’s return to one of the papers that kicked off this discussion, as it has a few examples. From page 810: “seroconversions occurred in three male and seven female contacts after enrolment in the study” (that’s not even counting whatever seroconversions had to have occurred in those subjects already HIV+ at the start of the study). There must be some explanation for those seven seroconversions.
Whether or not one believes HIV is sexually transmitted, seroconversions occurred. What were their causes, if not HIV infection due to sexual transmission?
Where does the HIV come from if not an external source?
I’m having difficulty squaring some of your comments on seropositivity and HIV testing with the high reported sensitivity & specificity of properly conducted HIV tests.
I do not see the Madrid study as “largely worthless”, though I’d give it less weight than the Padian et al. study for the reasons you give. (I should emphasize that I’m talking about the Padian et al. study as a whole there; I maintain my reservations about the zero seroconversions result.) Still, the Madrid study sample’s not particularly small compared to the number of Padian et al. couples who consistently had unprotected sex and were followed up prospectively.
Yes. The point I was making was that with the sequence analysis confirming the source of the transmission, the probability of any given seroconverter having been infected by IV drug use with strangers is less than if the researchers hadn’t done the sequence analysis (although of course still more than zero). So, to have me agree that the HIV infections are due to IV drug use, I would have to be convinced that the seroconverters had been sharing needles with their partners. And...
...as the study took place in Uganda, I find it very unlikely that even half of the seroconverters were infected by sharing needles (especially since none of the subjects reported injection drug use). Unless you are now arguing that seroconversion doesn’t actually indicate HIV infection (and I can’t tell whether you are or not), I expect you can see why I’d want alternative explanations for each of these seroconversions.
The African study, I’d say, is weaker evidence than the Padian study. At the same time, I think it’s absurd to write it off as “not useful”. At the risk of repeating myself, my reasoning goes like this: some number of subjects acquired HIV during the Ugandan study; exposure to HIV is necessary for infection; it is not credible to suppose that every HIV exposure that led to seroconversion is attributable to IV drug use. If there were only four or five seroconversions, I’d say it was statistically possible. In fact, the study reports that out of 239 monogamous couples that were initially HIV-discordant, 72 “acquired HIV during follow-up”. The probability that all 72 of these Ugandans acquired HIV by IV drug use is surely astronomically small.
Not especially, to be honest.
I’m not much impressed by Mullis’s Nobel credentials. There are quite a few Nobelists who got the prize and promptly started dabbling in crankery. (Including, as it happens, Luc Montagnier, so this isn’t just an issue for people who downplay the links between HIV & AIDS.) Jim Watson co-discovered DNA’s structure, but I’m not about to take all his proclamations about genetics at face value.
Anyway, I haven’t seen anything in the Uganda study report to suggest they used PCR in so sloppy a way as you’re suggesting, and I doubt they’d have much to report if they had. As such, it’s hard for me to avoid the feeling that your remarks about PCR are something of a fully general counterargument against HIV sequencing.
Right, but two samples of virus from a couple where one member infected the other will nonetheless be far more genetically similar on average than two samples of virus from randomly selected infected people in Rakai.
First off, before we get into anything else, we need to understand and agree on what seropositivity actually means. Seropositivity means a positive result on an antibody test such as western blot, to some combination of antigens identified by Gallo in 1985 in his large scale blood screening tests. He did a large number of tests on transformed (immortalized) cell lines derived from AIDS patients and found a combination of antigens that could screen blood—separating AIDS and pre-AIDS blood from regular blood.
The problem is he (like Motagnier) failed to isolate the ‘virus’, and most or all of the antibodies in the test react or cross-react with antigens to opportunistic microbes (candida namely) and cellular debris. The p24 protein in particular is essentially just a normal cellular wall or microvescile component—so the ‘HIV’ test is really just a general test of opportunistic infection and apoptosis or immune directed killing of CD4 cells (possibly due to widespread viral parasite burden). It is not a measure of ‘HIV’, it is a direct measure of declining CD4 cells and AIDS or pre-AIDS.
more on this
and a longer, more detailed analysis of cross-reactivity for the different ‘HIV’ proteins
The antibody tests are not standardized geographically or temporarily, so it also makes it very difficult to compare across studies—“seropositivity” means different things in different places and times.
As just one random example—most dogs typically have a mix of ‘HIV’ antigens, and are HIV ‘seropositive’ in whole or in part:
from this paper
You post a link to a paper which supposedly shows the ” high reported sensitivity & specificity” of HIV tests. This is not actually what that paper is about, but it references several other papers for this claim—the first I investigated being this. The important quote:
So they are just using Western blot as ‘confirmation’. So they are just using one antibody test to confirm another antibody test—which of course is rather ridiculous.
To actually compute the sensitivity and specificity for a “HIV” test, one needs a gold standard such as viral isolation or perhaps a DNA test. Unfortunately HIV can not be isolated, either because it doesn’t exit or it exists in only minute quantities.
But one can attempt to use the presumed viral DNA as a gold standard, and the result is extremely poor sensitivity and specificity:
Poor sensitivity, specificity, and reproducibility of detection of HIV-1 DNA in serum by polymerase chain reaction. The Transfusion Safety Study Group.
Poor sensitivity is perhaps a gross understatement—the study actually shows that around 18-25% of the population at large test positive for ‘HIV DNA’, and this is only weakly correlated with seropositivity.
You completely dismiss Mullis’s argument based solely on an ad hominem “not much impressed by Mullis’s nobel credentials” without seeming to acknowledge or understand the argument itself.
Seroconversion in the west is closely correlated with AIDS or pre-AIDS. This does not appear to be as true in Africa, so we are generally talking about two different worlds. Part of this may be genetic (black americans have amazingly higher seropositvity in general), the other part may relate simply to higher precedence of opportunistic infections and antigens that seropositivity measures. KS for example is vary rare in the west and along with systemic candidaisis was part of the original AIDS definition, but it is one of the most common cancers in Africa.
I am arguing that.
Seropositivity does not strongly correlate with ‘HIV’ infection (by DNA test), which is why it is better to discuss AIDS itself as being sexually transmittable or not.
The Gallo blood test is tightly correlated with AIDS (at least in the west) - simply because that is what it was designed to do, so you can use that as data for AIDS transmission discussions.
OK. At this point, I’m going to have to disengage and walk away from this debate. I’m realizing that the inferential distance between us is far bigger than I originally thought, and trying to bridge it would need me to considerably ramp up the effort I’m already putting into this. (Even then I can imagine this going on indefinitely, which isn’t a very appealing prospect to me, nor to other Less Wrong posters, by the looks of it.)
I’d still like to respond briefly to one part of your comment, which comments on my own words rather than HIV/AIDS:
It’s wholly legitimate for me to respond to someone citing Mullis’s credentials (as if I didn’t know about them already) by explaining why I give them little weight, and my next paragraph was meant to summarize why I gave “your remarks about PCR” (that is, those you paraphrased from Mullis) short shrift. In other words, I acknowledged the argument by rejecting it.
I’m glad you can walk away, I have a harder time initiating that. I’m curious though about the direction of the inferential distance you see—do you have a biology background?
The dissidents point to a rather surprising pile of evidence that the serological HIV tests are based on rather general, cross-reactive antibodies, and this is essentially a fundamental flaw in HIV science which has never been corrected. Now it may be that the orthodoxy has a really good counter to this, but if they do I have yet to find it. The orthodox position on this, from papers linked to wikipedia, points to studies which measure the sensitivity of various HIV antibody tests by comparing them to . . other HIV antibody tests.
The few large double-blind meta-studies that compare the different antibody tests to PCR tests show terrible sensitivity and specificity between the two, and I haven’t seen the orthodox counter to this. So something is wrong with the antibody tests, the PCR tests, or the whole thing. I imagine it’s a little bit of both—the antibody tests are cross-reactive (hence many dogs test positive), and PCR tests are difficult and subject to experimenter bias.
Perhaps the orthodox counter is that there are a whole big host of HIV related viruses, and the antibody tests are cross-reactive across these related species. This seems to then just beg more questions than it answers, and doesn’t circumvent some of the specific non-viral cross-reactions the dissidents point to.
My paraphrase of Mullis’s argument may actually be a mix of other dissident positions. I just rechecked that part of his book and he covers the difficulty of PCR and the confirmation bias but largely in regards to the OJ trial. On HIV he mainly rehashes Deusberg’s argument.
All right, enough.
None at all. (I expect the inferential distance would be even greater if I did. If I had personal experience of working with retroviruses, for instance, I reckon my prior probabilities for claims like “HIV can not be isolated” or “HIV doesn’t exist” would be far, far less than they are. And they are already very low.)
So those electron microscope pictures are fakes?
Which electron microscope pictures?
When ‘HIV’ was first ‘discovered’ in the original papers by Gallo and Montagnier, they had difficulty isolating and didn’t publish pictures from what I understand—that didn’t happen until years later. Gallo’s great discovery for HTLVIII was based on running a lager number of antigen/antibody tests with an immortalized cell line to find an antibody test that could screen AIDS and pre-AIDS blood from regular blood. That is the basis of all the current HIV tests.
The first published pictures came more than a decade later, and they showed that “HIV isolate” really consists largely of cellular debris and microvesciles. In these EM photos, they do find some occasional particles of roughly the right size and label them as “HIV”, but they could also just be any of a number of other things, and for all intents and purposes, HIV ‘particles’ look like regular microvesciles.
The titles of the papers say it all:
“Cell membrane vesicles are a major contaminant of gradient-enriched human immunodeficiency virus type-1 preparations”
“Microvesicles are a source of contaminating cellular proteins found in purified HIV-1 preparations”
more on HIV ‘pictures’
I have lost the link, but there are better more recent pictures taken with ATM, and they show that for all intents and purposes, it’s impossible to distinguish ‘HIV’ from regular microvesicles that bud from the cell wall naturally. If HIV can be said to exist at all as a unique exogenous virus, it is only because of unique RNA content in the microvesicle, and in this sense is very much unlike all other known viruses.
Of course, the part of HIV’s genome which is supposed to code for the outer envelope is pretty much the same as the endogenous sequences that already exist in the human genome—the HERVs.
Huh. While I still think that the HIV explanation is the most likely one for AIDS, I am slightly less convinced.