You selfish fuck! (Man with possible HIV cure in his veins refuses to undergo tests)

The BBQ Pit
41

Sorry for the tripple post, but I’m thinking maybe we got off on the wrong foot?

I do believe that a simple driveby wiki cite isn’t sufficient proof. You can’t really fight ignorance if you’re not even going to take the time to explain things, and if when asked to explain them you simply respond with insults. I may very well be ignorant about advanced probability mechanics, but it is disingenuous to claim that they’re basic. I’ve taken several courses dealing with stats, and never once touched on anything that you cited. I’m aware of the Gambler’s Fallacy, and how individual tosses of a coin are 50/50 no matter how many have come before it, etc… Now, if you’d like to fight whatever ignorance I, and I’d wager, many people reading this thread have, then please do.

Want to start over?

42

First off, sorry to the OP. If you wish for me and others to take this hijack elsewhere, I will, and apologize accordingly.

However, I think it is at the very least relevant to our discussion, so I shall continue.

I think I see where the problem lies. When 10,000 people are tested, 2 actually test positive. One of these people have HIV, and one does not. The one who tested positive but does not have HIV is the false positive rate. The one who tests positive and has the disease is the sensitivity.

It is sort of hard to see, but here is a pictorial representation.

Do you see where I am coming from now?

43

Have you really never seen what he is talking about. It is very basic in probability.

If you have a test that has a 1/1000 chance of false positive and say 1/100 chance of false negative you can construct a table.

There is one other piece of information the distribution of the disease in the population. Lets assume 1/200.

assume 100,000 people get the tests.

There are a total of 500 people with the disease.
There are a total of 99,500 without the disease.

99 people with test positive without having the disease.
495 people will test positive with the disease.
Therefore if you test positive you have a 1 in 6 change of having the disease not 1 in 1000.

These number change a lot with how common the disease is. As the disease becomes more common in the probability that a positive test means you have the disease goes up.

44

I’m relatively confident that you’re misunderstanding the level of false positives from an antibody test. They are extroirdinarily sensitive and accurate, and probably less than 1 in 10,000 positive test results are wrong. Not 1 in 10,000 total tests administered.

45

Nope, never.

And thanks for the explenation, but I still can’t understand how if the test is 99% accurate for any positive result why there isn’t a one percent chance that you got a false positive.

I don’t understand why they have to be taken as a group instead of isolated incidents.

How does it differ from the probability of a coin toss?

46

I’m a bit confused on that false positive probability stuff myself.

47

This comes about because there are so many people without the disease that there is a large pool of false positives compared to actual positives.

48

There are natural frequencies and there are probabilities. A natural frequency is a ratio, say 1 out of 100, where as a probability is a percentage, like a 1% tax. It is relatively difficult to derive understanding of data using probabilities, since we are not wired to think in that way.

For instance, “about 0.01 percent of men with no known risk behavior are infected with HIV. If such a man has the virus, there is a 99.9 percent chance that the test result will be positive. If a man is not infected, there is a 99.99 percent chance that the test result will be negative.” (From my cite above.)

Now, what the hell does this mean? Most believe, given the above, that if you test positive then you have a 99%+ chance of having the disease. This is not the case, but is a very normal, very natural way to think. It is innumeracy, plain and simple.

Maeglin cited Bayes’s rule, which is entirely correct. You take the above and apply his rule, giving you some more easily digestable data. However, you can just as easily use some common sense to derive a natural frequency.

So, try to “mentally transform” the above. Out of 10,000 people, then, 9,999 will not have HIV. One will.

Out of 10,000 people, 2 will test positive. One will test positive with the disease, and one will test positive without it.

Those that test positive but don’t have the disease is 1 out of 10,000. This gives you the false positive rate from above. However, this does not mean you have a one percent chance of recieving a false positive.

The chance of receiving a false positive is a ratio, and accordingly is a part over the total. In this instance, the total is the total number who tested positive. The part is the people who tested positive but who do not have the disease.

From our above example, 2 people have tested positive, and accordingly this is the total. Only 1 person actually has the disease out of these 2 people, so 2 - 1 gives us our part, or 1. So, the chance of recieving a false positive is 1 out of 2, or 50 out of 100 or 5,000 out of 10,000. Any way you cut it, it is still 50%.

49

First, what is the difference between tests administered and test results? The results of those administered tests are what we are using, unless I am missing something.

Second, minus the “probably” part, you are entirely correct in saying that 1 out of 10,000 tests is a fasle positive. This does not mean that the test is 99.9% accurate, however.

As stated above, but hopefully stated here more clearly, you have a 1 out of 10,000 chance of having HIV and recieving a positive result. You also have a 1 out of 10,000 chance of reciving a false positive, which is a positive result despite not having the disease.

This gives us 2 positive test results out of 10,000.

The accuracy of these tests is measured by taking the false positive rate and dividng it by the total number of positive results. Since we have 2 total positives and 1 false positive, the accuracy of the test is only 50%.

50

We probably did.

This is a little tangential to the issue at hand, but since others above have already explained it, I think this is worth exploring.

The wiki cite is not “proof”. It is not an argument from authority, nor was it meant to be particularly mysterious. Bayes, and later Laplace, did prove how to solve these kinds of problems. Whether I cite wiki or math.com or some random person on the street, the proof remains the same. The proof of Bayes’ Theorem is absolutely trivial: you can find it anywhere and armed with a half hour of reading, you would certainly understand it.

This is absolutely true. But I also expected you, a person interested in educating himself, to have given the cite a reasonable perusal before complaining that it was a wiki and that I had failed to walk you through it. If I had cited someone’s dissertation in cohomology with no further remarks, then I would be an unhelpful asshole. But it is not disingenuous to claim that this is really basic stuff, and that intelligent people who take the time to check it out usually grasp its intuitiveness very quickly. Most people, more or less, think in accordance with Bayes’ Theorem, so it is often exciting for them to see it formalized.

I do confess to some surprise that you have never come across Bayesian inference at all in statistics. It’s just something that tends to come up, even if you are not getting trained as a Bayesian.

I interpreted your response, perhaps wrongly, as being extremely lazy. I deal with lazy people unwilling to self-service all the time at work, so I snapped back a little too harshly.

Sure. If the above responses haven’t satisfied you, I’ll take a crack at them tomorrow. Or if you want to take it off-board, feel free to shoot me an email.

51

Maeglin: thanks for taking the time and being civil, I really appreciate it. I’ll chew over what’s been posted, but I did feel before this as if I was rather well versed in stats. I had always thought that, for instance, any individual coin toss was 50/50, but the probability of ten in a row being heads was a different matter. As often happens, it seems I may’ve been ignorant as to what I was ignorant of.

I’ll see if I can’t make heads or tails (pun unintended) of this stuff, and I’ll repost or email ya lata. Thanks again.

52

The difference is between tests administered and positive test results like in my prior post. For instance, if a test is totally accurate, you have a 1% natural frequency of a disease, and you test 100,000 people, you will have total of 1,000 positive test results and 100,000 test administrations.

What I’m saying is that the test is extroirdinarily accurate even among the positive test results. I understand your argument about the proportion of positive test results versustrue positives, but in real life the means through which AIDS tests are conducted mean that fewer than 1 in 10,000 positive test results are false positives.

Also, let’s reiterate that there were no lab errors in the false positive. The samples which gave the “false positives” the first time were retested multiple times in 2002 and 2005 and consistently gave positive results for HIV antibodies in this man’s blood. The question is how they got there. The most likely source, whenever the test “works” is when you have an HIV infection. However, there are apparently cases in the medical literature in which HIV antibodies or at least antibodies which this HIV test is sensitive to have been found in the absensce of HIV infection among certain populations such as people that have had recent flu shots, etc.

Let me try to make an anology that’s a bit clearer. In addition to a lot of other clinical tests, testing for sugar in the urine is a commonly used means to test for diabetes. Because of poor blood sugar management, patients can become so hyperglycemic that sugar leaks into the urine. So, we have some sort of clinical test that looks for sugar in the urine and therefore diabetes. Let’s say for the sake of argument that this test for the presence of sugar is 100%, always, totally, say never, irrevecobly accurate. This is close to how accurate the HIV antibody tests are. However, again for the sake of argument, let’s pretent that there’s some very rare group of sexual deviants that gets off on sticking themselves with IV’s and pumping themselves full of 50% dextrose solutions until they get blood sugar counts in the 4-500’s (typically for an unmanaged diabetic, 80-120 is normal). Of course, sick puppies that they are, these people don’t actually have diabetes, but they do have sugar in their urine. So, of course our 100% infallable clinical test will still show that they have sugar in their urine, and we’ll falsely conclude from that that these patients have diabetes, even though they don’t so the test will have a, “false positive”. That is the most likely explanation for Mr. Stimpson, and similar, “false positives,” are described in the literature, but I’m open to other explanations. This seems more likely than a spontaneous cure.

53

[hijack]
Re: the HIV immunity thing
Yes, it’s true. Yes, no-one tells you. No-one knows because a little bit of knowledge about a fatal cantagious virus is the most dangerous thing.
It is not a recent discovery, they’ve known about it for at least 10-15 years. Turns out it’s probably from the plague, people who survived survived because of the mutated receptor.
From the stats I read 1 in 100 are immune, 1 in 10 carry one allele of the mutation. That is, if you’re family is descended from plague ravaged areas.

And yeh, I brought cites.

The thing that annoys me most about this is that people think this is recent development. Go to your local library, find a specialist book written about HIV from the 90’s, it’ll be there. It’s not AMAZING or NEW, it’s evolution.

It is not a cure. Unless you want to start genetically engineering people, this won’t help.
[/hijack]

54

I’m sure that you know this, gazpacho, but in the interest of fighting ignorance, it’s worth noting that “population” in this case refers to the group of people tested, not necessarily to the general population. The frequency of HIV in people seeking out tests for HIV is probably higher than that in the population as a whole.

55

Wow, it’s weird that both Bubonic Plague and HIV immunity are mediated by mutations on the same receptors. Very interesting stuff, I thought the prevalence was like 1 in 100,000 not 1 in 100 for immunity.

Unfortunately it still doesn’t explain this case, but it is interesting.

56

I think, ZebraShaSha, you’re missing something. Sensitivity, in epidemiological terms (which is what we’re referring to in this case) means the proportion of truly diseased persons in a population who are identified by a screening test. Or rather it is a measure of the probability of accurately diagnosing a person with the disease. Specificity means the proportion of truly non-diseased persons in a population who are so identified by a screening test.

I took this pretty much verbatim from A Dictionary of Epidemiology, ed. JM Last, because I get tongue-tied trying to explain it myself. What this means is that if the ELISA test, according to FinnAgain’s cite, is 99.5% specific, it means that for any positive person tested, there is a 99.5% chance to get the correct result. The reason the cite further notes that the specificity is slightly higher is because of the number of people who are negative but test positive will be slightly higher because of the sensitivity.

It’s calculated simply (to borrow from Last again):

Screening test result True status Total

                                          Diseased      Non-diseased

Positive a b a + b
Negative c d c + d
Total a + c b + d a + b + c + d

a = true positives
b = false positives
c = false negatives
d = true negatives

So sensitivity is expressed as a/a + c and specificity is d/b + d
The predictive value for a positive test result is a/a + b
The predictive value for a negative test result is d/c + d

So your example with the 10,000 people is entirely dependent on how many people in a given population actually have the virus, which can then be compared with the results.

That said, no reputable lab will ever do only an ELISA test. It is always followed up with a Western Blot to confirm, which virtually eliminates false positives completely.

57

Wow, my table didn’t turn out well at all. Anyone know how I could fix that?

58

I am not sure about the veracity of that figure, and I have no cite for it.
Why is it weird? They both enter the cell the same way. Therefore, a mutation in the receptor makes you immune.

59

Oh, and, it might. The HIV could still be in the blood for a certain amount of time, and just not able to enter his cells, so it eventually dies.

60

Right, I understand that, I just found it odd that they happened to use the exact same receptor, after all, aren’t there dozens if not hundreds of analogous cellular receptors for endocytosis that HIV could have just randomly chosen (or will choose) instead?

Also, I suppose that if this individual had both a natural immunity from a CD4/CCR5/CXCR4 receptor mutation and exposure to a very large number of HIV viruses from whatever then that might produce an immune response and a positive antibody test result, but I just cannot imagine this happening from sexual contact. Maybe IV drug use if he like took in an entire ml of HIV virus laden blood from flashing or whatever.