Friday, May 7, 2010

Does phlebotomy 'work'?


There's a discussion in a nice book about the history of Islamic science (Ehsan Masood, Science and Islam: A History) of a man named al-Razi, who in about AD 900 was said to have done a carefully controlled experiment to test whether phlebotomy (blood-letting) worked as a treatment for meningitis. Some patients were given the treatment and others were untreated 'controls'. Al-Razi found that the bloodletting worked, in that more of the treated patients than controls recovered.

This therapy was part of the ancient and revered view of life upon which the classical medical approach codified by Galen was based. The humoral theory, that existence and hence life and health are based on balance of four basic properties (earth, air, fire, water), that in humans corresponded to blood, black bile, yellow bile, and phlegm.

Everything could be explained in terms of disease as the state in which these are out of balance. Blood-letting was done when the patient was deemed to have an imbalance by an excess of blood. Galenic medicine lasted for many centuries and it was verboten even to question it. And why question it? It worked! That is, some patients got better and the belief in the system led everyone to accept its sometime success as supportive evidence (and indeed, it's possible that even when assessed by modern scientific standards, bloodletting may sometimes have done some good, as this story describes).

Why don't we accept it today? In fact, even al-Razi himself wrote a book casting doubt about the degree to which Galenic medicine was true. After all, we accept modern medicine even though it fails to cure everyone. We have to ask what causation really is. After all, placebos work. If you know people are praying for you, it apparently works -- although prayer doesn't work if you don't know people are praying for you.

We dismiss that as 'only' psychological, even if that is purely physical and molecular, by involving neurotransmitters that affect other cell behavior, such as by the immune system and who knows what else, eventually leading to improvement in the disease. Blood-letting apparently has a measurable, replicable physiological rebound effect that makes people feel better a few hours later. We say these things don't really cure the disease, or if they're just psychosomatic, somehow that doesn't count. But if the brain is a material rather than immaterial structure, and the effect is thus material, why doesn't it count?

We want higher percentages of success. We want therapy to be direct, rather than indirect. If the treatment is believed by the patient, it boosts his immune system in some way, etc. Somehow, targeting the true pathology indirectly, rather than by targeting the proximate molecular cause, is not considered 'real'.

But that's our own culturally derived way to define medicine and its efficacy. It's similar with diseases like AIDS and HIV. As the South Africans said for a decade or more, poverty is the true 'cause' of AIDS, not the virus. Unfortunately, many thousands died as a result. Yet poverty is still causally associated with HIV infection. South Africa has finally accepted that HIV is also a cause of AIDS, and thousands or millions of lives may now be saved as a result.

Empirically, the desired explanation can be chosen to be some net result -- 'cure' in the case of disease. Science in the west, at present, wants reductionist molecular explanations, about proximate cause. Causes higher up the material chain -- like poverty and poor education cause poor neighborhoods with no good grocery stores cause reliance on McFastFood causes obesity causes high blood pressure or glucose causes retinal and peripheral neuropathy causes blindness and loss of extremeties. So what causes blindness? Even in our molecular, reductionist, technical age, diabetics still become blind.

There is no one answer. If removing poverty greatly reduced blindness, isn't poverty a cause? Or McBurgers? The prevailing view is that if we identify some ultimate cause -- the preferred target for many in science these days is your 'personalized' genome -- we will get to the 'real' cause and will then live forever. But the focus on genes is part and parcel of the structure of our current society.

Whether one approach to causation will ever, by itself, lead to miraculously high levels of efficacy nobody can say. Galenic physicians thought they had the ultimate answer. Collinsian medicine (Francis Collins, Director of NIH and the chief spokesperson for personalized genomic medicine) is having its day today. What about tomorrow?

The same kinds of questions arise in evolutionary and developmental biology. We've recently posted on phenogenetic drift-- the idea that essentially the same trait can come to be due to different genetic bases even while being conserved by natural selection -- which suggests that genes contribute but are not 'the' cause of the trait. This is related to the entire concept of complex causation.

So was al-Razi right that phlebotomy cured meningitis? Perhaps it is inappropriate to ask whether Galenic medicine 'works'. It is more interesting, to us at least, to ask what we mean by 'works'.

[p.s., al-Razi, known in the west as Rhazi, wrote critically about Galenic medicine in a book Doubts About Galen]

21 comments:

James Goetz said...

I suppose that the bloodletting worked better than no treatment while a placebo would have worked better than the bloodletting. And I'm not mocking psychosomatic illness. And for ethical reasons, I'm also not promoting placebo prescriptions.

Ken Weiss said...

Maybe psychological therapy, based on what we know about the placebo effect, needs to be investigated and tried more intensely, pulling back from so much direct kinds of intervention.

An important question is how to use 'psych' therapies without spoiling their efficacy if patients have to be deceived, as those receiving placebos in clinical trials are. This must have been studied by people before.

James Goetz said...

Perhaps encouragement in thinking positively about our bodies ability to repair itself could help. I suppose that their could be a controlled study with one group given a placebo while another group is encouraged in such positive thinking. I would hope that there would be no significant difference between the groups, but I suppose that there's a chance that the deceived group would do better.

Various theists can benefit from overcoming psychosomatic illness through prayer. I understand that philosophical materialists would say that any positive effects of prayer are nothing but a placebo effect. I disagree with that while I agree that the placebo effect has its place in prayer. But this approach wouldn't help, for example, non-theists unless they're questioning their non-theism.

Ken Weiss said...

I believe that a recent experiment showed that if a patient (I forget what disease this concerned) knew s/he was being prayed for, there was a benefit. But if they didn't know they were being prayed for, there wasn't. What faith or worldview the patients had is something I can't remember.

There are brain centers that at least some believe not only participate in religious or meditative experience, but that can be used by magnetic flux to induce such experiences. If this is correct (the research has been challenged), then there may be ways to induce the relaxing or perhaps even rapturous experiences that could be good for well-being and, who knows, also perhaps improved response to disease.

Zachary Voch said...

I remember reading a paper by Graham Oppy on causation (though I can't remember the title) about the problem of preemption when addressing causation. The example being: Suzy throws a rock and immediately afterwards Jack throws a rock. Suzy's rock smashes a bottle, right after which, Jack's rock passes through the location where the bottle previously rested.

Now, the `proximate cause' in this situation would be Suzy's rock. But, if we describe causation counterfactually, `if Suzy had not thrown the rock, the bottle would not have broken,' we would have a false statement as Jack's rock would have done the trick (barring the addition of complicating circumstances to our hypothetical).

So, take a patient X who is both poor and has AIDS. We can examine two statements:

"If X was not poor, X would not have AIDS."

"If X had not contracted the HIV virus, X would not have AIDS."

These aren't contradictory, so we can compose them:

"If X was not poor, X would not have contracted the HIV virus implying X would not have AIDS."

Of course, this statement is not necessarily true, since not only the poor develop AIDS (as it is also true that AIDS is not only due to HIV, but we ignore these cases). But, poverty alone is not a vector for HIV, so we have to complicate this a bit:

"If X was not poor, X would have been able to easily afford condoms. Since X could easily have afforded condoms, X would have bought them. Since X bought them and knew that condoms prevented pregnancy/some stds and wanted to avoid those outcomes, X used condoms. Since X used condoms, X had very little chance of contracting HIV. Since X did not contract HIV, X did not develop AIDS."

Now, all of those claims in this logical chain are empirical and have been (or could be) studied empirically. And indeed, many have been, not just at the molecular level. Many of these, like the `what causes blindness' scenario, include parts of the chain that are only open questions in the remotest sense of the term. Fast food causing obesity and poverty increasing consumption of cheap food like fast food are hardly mysteries for science to tackle. It is hardly arcane or contentious knowledge that poverty predisposes the population to many diseases, not just diabetes or HIV. How to develop safe and effective drugs that prevent blindness once diabetes has already been developed, however, remains an open question. Similarly, how to vaccinate against HIV remains an open question.

Barring other proximate causes of AIDS, as HIV is the major one, stopping HIV stops AIDS. Preventing poverty alone is not enough. The best possible method of reducing AIDS is not fighting poverty, but developing a vaccine. At the same time, poverty should be addressed, but as far as a targeted cure or treatment for AIDS is concerned, elimination of HIV is paramount. Still, until that occurs, we should address what causes we can, like poverty, instead of sitting on our hands. The molecular reductionist approach seeks to end the battle, but we can still treat the wounded and prevent many casualties in the meantime.

Zachary Voch said...

However, to preempt the development of AIDS or blindness at the higher level of the causal chain has gone beyond the questioning of the material sciences, which focus on (and in my opinion, should focus on) the open questions at the molecular level. The implementing of programs to reduce poverty are questions of policy and sociology. Science can tell us the treatment of these causes, if one cares to look, but it can not force policy-makers to implement social programs to help families afford better food. Though the biologists and chemists might be busy with reductionist methods, we still have psychologists and sociologists to answer questions like `how do we increase the usage of condoms?'

As far as the problems on the higher chain, the problem isn't one of a simplified view of causation, but rather the lack of implementation of solutions which we already know empirically. A Galenic science isn't the problem here.

That being said, we should take known facts about placebos into account in treating patients. Some placebos are stronger than others, environment is important, etc. If the goal is to treat or cure, placebos are a known assistant. The main reason to not call a placebo itself a `treatment' of a disease is to keep the term from becoming a triviality. For then yes, anything is a `treatment,' but this confusion of language would do little but further confuse patients or risk equivocating `Placebo +' and `Placebo.' Nontrivial treatment is of the form `Placebo +.' That's not to call trivial treatment `trivial' in the sense of being unimportant.

Also, distrust of doctors due to dispensed placebo treatment would cause a `nocebo' effect, which could risk hurting patients and decrease the effectiveness of a remedy in the same way a placebo increases its effectiveness.

Ken Weiss said...

These are very thoughtful comments. I think nobody (sane) would doubt that attacking HIV by a vaccine would interrupt the causal chain at an important point. Interrupting poverty would, if I understand you (and I'd agree) interrupt the causal chain at a higher level. That's more remote from the more proximate (HIV) cause.

But stopping poverty might not stop the actual diseases of AIDS, which can arise in the absence of HIV (though HIV greatly raises their probability).

HIV denial is denying what appears a clear-cut cause, and in that sense is very misguided. But the issue we were raising was about various concepts and types of 'causation'.

There was a good precedent, well-analyzed by Ludwik Fleck in the 1930s, of the ideas about syphillis and causation.

Perhaps a bottom line point is simply to be more sophisticated about notions of causation. We do molecular genetics in our lab every day, so we would not want to overlook that important aspect of these various issues.

Zachary Voch said...

"The same kinds of questions arise in evolutionary and developmental biology. We've recently posted on phenogenetic drift-- the idea that essentially the same trait can come to be due to different genetic bases even while being conserved by natural selection -- which suggests that genes contribute but are not 'the' cause of the trait. This is related to the entire concept of complex causation."

I take it that you are referring to http://ecodevoevo.blogspot.com/2010/04/ant-misbehavin.html as the recent post. With genetic drift, the *stability* of a given trait is due to natural selection, as you noted, but the trait itself is still a product of genetic mutation. If `genetics' is taken as a broad term for both the gene(s) of a given trait and those regulating the trait, we still have genetics as `the cause' of existence and expression of a trait, and selection processes as `the cause' of the stability of the trait. Given, these are categories, not single occurrences, but what is left unstated as a given are environmental necessities not particular to a trait, which returns us to our placebo case:

"We say these things don't really cure the disease, or if they're just psychosomatic, somehow that doesn't count. But if the brain is a material rather than immaterial structure, and the effect is thus material, why doesn't it count?"

Because the material effect is not consequent to the material of the treatment. A sugar pill causes a material effect in the brain leading to improvement, certainly, but it does it through the reaction of the patient. In other words, `patient taking sugar pills' is the treatment, not `sugar pills'. In the same way, `patient supported by family and community with prayer' is the treatment as opposed to `family and community communicating with higher power' being the treatment. The distinction between the two is an empirical one, determinable via experiment.

Lumping higher level causes, like poverty, under the term `cause', confuses these distinctions. `Poverty causes AIDS' and `Poverty correlates with AIDS and indirectly predisposes a population to developing AIDS' should be kept as separate statements. Else, it would be like saying that Suzy's parents broke the bottle. Sure, at some level, they did in that scenario. However, science does not need to focus on questions like "can sex result in reproduction?" in addressing the problem of broken bottles. Rather, material scientists should focus on developing stronger bottles.

Zachary Voch said...

@Ken Weiss

Thanks for the feedback :D

"Interrupting poverty would, if I understand you (and I'd agree) interrupt the causal chain at a higher level."

To be more precise, I would say `decrease the incidence of AIDS', as causal chains for AIDS could still be constructed without poverty. However, poverty adds another logical pathway to AIDS.

"But the issue we were raising was about various concepts and types of 'causation'."

True, I tried to touch on particular aspects of your examples, and in doing so, I didn't focus entirely on causation.

"Perhaps a bottom line point is simply to be more sophisticated about notions of causation."

Perhaps you could inform me if the case is otherwise, but from what (little) I know, I'm under the impression that items like `poverty correlates with higher incidence of diseases like HIV' to be common knowledge, at least among the scientific community. For this reason, I stated that "As far as the problems on the higher chain, the problem isn't one of a simplified view of causation, but rather the lack of implementation of solutions which we already know empirically. A Galenic science isn't the problem here."

Though as a whole we need to address problems like AIDS on every possible level, those who are involved at the molecular reductionist level are doing exactly as they are supposed to be doing. As we've other areas of expertise that focus on the social aspects which are quite outside technical aspects on which biochemists are experts, I thought it was safe for them to focus on the molecular level of causation, especially given the awareness of higher level correlates. If as a population we did not know that obesity was caused by fast food, by all means, biologists should let us know.

But again, I don't think that unsophisticated views of causation are a problem, at least as far as the examples that you provided. There are plenty of examples I can think of where oversimplicity or even an absurd monocausality is a problem, but these lie in politics and other areas, not in science.

If you have other examples, I would be interested.

Anne Buchanan said...

Zachary,

Determining 'cause' is problematic enough, and determining at which level to intervene to interrupt the cause-effect pathway can be hugely problematic as well. But that has a lot to do with what we mean by cause.

You say: "But again, I don't think that unsophisticated views of causation are a problem, at least as far as the examples that you provided. There are plenty of examples I can think of where oversimplicity or even an absurd monocausality is a problem, but these lie in politics and other areas, not in science."

Here you hit one of our hot buttons! I would suggest that there are certainly unsophisticated views of causation in genetics, epidemiology, anthropology, psychology, and so on -- the idea of single gene causation for complex diseases and traits is still rampant, despite the overwhelming evidence (which has in fact been true since the early 1900's) to the contrary. Not to mention the widespread idea that natural selection is the cause of the evolution of most traits.

But often the problem is not one of unsophisticated views of causation but more one of how causation is determined in the first place. What are the standards for determining the cause of infectious disease? The Koch postulates, which most people, including Koch in 1890, agree can't all always be met, even if a microbe actually is determinative.

--The microorganism must be found in abundance in all organisms suffering from the disease, but should not be found in healthy animals.
--The microorganism must be isolated from a diseased organism and grown in pure culture.
--The cultured microorganism should cause disease when introduced into a healthy organism.
--The microorganism must be reisolated from the inoculated, diseased experimental host and identified as being identical to the original specific causative agent.

How do epidemiologists determine the cause of a chronic disease? The best the field has agreed upon are the Hill criteria (too much to reproduce here, but readily Googleable), but again, only one (cause must precede effect) can be thought to actually be required.

Determining causation for complex traits is tough -- for methodological and epistemological reasons. This is why we don't know seemingly simple things such as whether coffee is good or bad for us, whether we should eat butter or margarine, or eggs, what causes heart disease, or which genes 'cause' face shape or eye color. We know genes don't work alone, so can we identify causative gene networks? But eventually, every gene in the organism can be included in a 'causative' network, so how do we decide where to stop? And so on.

Does this mean there is no true molecular cause for disease? Well, it depends on how you define cause. Is PKU a genetic disease or an environmental one?

Zachary Voch said...

@Anne

"Here you hit one of our hot buttons!"

Good! That's what I was shooting for.

"I would suggest that there are certainly unsophisticated views of causation in genetics, epidemiology, anthropology, psychology, and so on -- the idea of single gene causation for complex diseases and traits is still rampant, despite the overwhelming evidence (which has in fact been true since the early 1900's) to the contrary. Not to mention the widespread idea that natural selection is the cause of the evolution of most traits."

I agree that "the idea of single gene causation for complex diseases and traits is still rampant, despite the overwhelming evidence (which has in fact been true since the early 1900's) to the contrary" is a problem. We need to join these two parts to form another point on which we agree:

1) "Not to mention the widespread idea that natural selection is the cause of the evolution of most traits. "

2) "But often the problem is not one of unsophisticated views of causation but more one of how causation is determined in the first place."

But where we agree, we are not establishing that models of causation considered are too simplistic, but rather that a specific cause is attributed to something (like a selection process X causing a trait A) without sufficient evidence or even reflexively. There, you don't have to convince me.

Going back a bit:

"...the idea of single gene causation for complex diseases and traits is still rampant, despite the overwhelming evidence (which has in fact been true since the early 1900's) to the contrary."

Here again we agree. As in your PKU example, a single gene can not be given entirety of credit for causing the disease (in all cases). But, I think you go a bit far with this:

"Does this mean there is no true molecular cause for disease? Well, it depends on how you define cause. Is PKU a genetic disease or an environmental one? "

If you define cause in terms of `absolute' and `always' and `singular', you will get a `no' answer, as would be true for anything I can think of, not just PKU.

In the case of (certain strains of) HPV and genital warts, HPV might be asymptomatic in some patients and not in others. But this does not imply that genital warts does not have a 'true molecular cause'.

HIV does not develop into AIDS in a certain percentile of the population, does this mean that AIDS isn't caused by HIV? We can stretch the term `cause', but as I stated before, we quickly mislead whenever we do.

In all of these cases, as in the ones you gave (healthiness of margarine, eye color, etc.), we have balancing factors. We might say `sunlight increases temperature of an area', but we are obviously excluding the case when sunlight is powering a solar panel which is generating electricity for an air-conditioning unit. So with PKU, we have a genetic cause and environmental factors. With Wilson's disease, environmental factors affect severity or even whether the person with the inherited trait has symptoms (increased copper in diet).

So of course, statements about causation are simplified, but I don't think that it's generally a consequence of an oversimplified view of causation.

The trouble with empirical statements about causation is where a cause is attributed without separating that cause from the complicating framework. As you stated, that is extremely difficult to do in many cases, so views about a given cause must be held only very tentatively.

Still, this is different from the point I was making to Ken Weiss. Here, we have definitive reasons for avoiding single gene explanations for many conditions on the molecular level. I was referring explicitly to `non-mechanistic' levels of causation and whether or not they are ignored/should be focused on by the technical community.

Anne Buchanan said...

Well, we'll just have to disagree on whether oversimplified views of causation are rampant -- and problematic -- or not.

So, before WWII, type 2 diabetes was essentially non-existant in Native American populations. Now it's endemic and Native Americans seem to have a fairly simple genetic susceptibility (not yet identified). But of course the genetics haven't changed in the last 60 years, but diet and exercise have. What's the 'cause' of the epidemic? And how do you best prevent or treat diabetes in this population? Would it help at all to find the genetic 'cause'?

I guess I come down to what the purpose of your search for cause is anyway. If you want to prevent AIDS, condoms can be one way, but needle exchanges or encouraging monogamous relationships, say, are others. If your purpose is to identify the _molecular_ cause, the answer is HIV. Of course knowing that you ultimately want to prevent HIV infection is crucial but doing that requires a lot of broader contextual knowledge other than about the virus per se.

How do you prevent diabetes? The quick and dirty answer is diet and exercise, no matter the genetic background.

But I think we all basically agree, and are essentially just fiddling with the issue of why we're looking for cause anyway.

Ken Weiss said...

Perhaps a summary of the main points of this discussion is that causation can be complex and involve multiple factors, none acting strictly on its own. Sometimes, this is only a trivial issue, but sometimes, if one wants or hopes to find single factors with very high causal predictability on their own, or to understand causation in its entirety (if that's ever possible), then we face challenging issues.

The drive to find single or simple causal explanations, when there is reason that we should be more comprehensive, is where we see problems. And there is variable use of the word itself ('cause') that sometimes means the ultimate molecular connection, and other times refers to what would be removed and would remove the consequence.

In this latter context, we earlier referred to papers by Ken Waters, philosopher of biology, and he deals with this in the context of genetics now, and in its early days.

Zachary Voch said...

So, addressing D(P) decreases the general incidence of diabetes, but addressing the genetic susceptibility of Native Americans will (less significantly) lower the general population diabetes rate and also equalize the incidence of diabetes under controlled conditions for diet. So yes, it would help to find the genetic risk factor, but at the same time it is clear that tackling the more immediate environmental cause D(P) reduces diabetes across the board and more significantly.

But, taking D(P) to be a clearly-established cause of type 2 diabetes, addressing the problem lies, once again, on the PR/political side and not the research side. The primary thrust against diabetes should be in reducing D(P)/obesity, but the general health issue of obesity is fairly well known within the public, and not just for diabetes. P knows (or could easily find out in greater detail) what the risks are associated with obesity, and P is the only party capable of decreasing D(P). The research has done its part on this front. The open problems then for research are items like preventing blindness in those who are already diabetic or determining/treating the genetic mechanism at work in predisposing higher-risk groups like Native Americans. Again, simplified views of causation among scientists are not the problem here. The open questions for addressing D(P) are sociological, not biological.

Zachary Voch said...

"So, before WWII, type 2 diabetes was essentially non-existant in Native American populations. Now it's endemic and Native Americans seem to have a fairly simple genetic susceptibility (not yet identified). But of course the genetics haven't changed in the last 60 years, but diet and exercise have. What's the 'cause' of the epidemic? And how do you best prevent or treat diabetes in this population? Would it help at all to find the genetic 'cause'? "

This is a great example. Here, we have a genetic predisposition/susceptibility with an environmental cause, whereas poverty is a predisposing factor for AIDS with HIV the mechanistic cause, we have a sort of role reversal in the case of diabetes in Native Americans.

So, let's consider a short definition of type 2 diabetes: "acquired insulin resistance." What causes this? `Acquired' isn't a part of `insulin resistance' so much as a linguistic presupposition about the (nature of the/time scale of the) cause of the `insulin resistance.' Within the language itself, we are already assigning primacy to an environmental cause, in this case, diet.

"type II diabetes" := "acquired insulin resistence" = "state of insulin resistance relative to general population due to environmental factors"

With the genetic susceptibility of a subset of the population in mind, we continue:

Let P:= "general population" and N:= "Native American population (more precisely, population with the genetic susceptibility)" and for x in P includes N, let A(x) denote the property "average incidence of type II diabetes" and D denote mechanistic causes (primarily sugar intake). So in a simplified scenario, D(P) could be "average sugar intake of general population."

So the claim of susceptibility is as follows:

If D(N)=P(N), then A(N)>A(P). That is, if dietary factors in developing type II diabetes are controlled, we expect a higher incidence of the disease in Native American populations than the general population. We would say the case of this implication alone is genetic susceptibility. Now, the implication

If D(P) increases, then A(P) increases

is the claim that increased average sugar intake correlates positively with increased rates of diabetes. So if we want to fight diabetes, we have a two-part battle: the first is the one that no doctor ignores (to my knowledge), that reducing sugar intake (particularly among the obese) will decrease the incidence of diabetes:

Decrease D(P) and D(N) to decrease A(P) and A(N)

Secondly, we might identify and correct the genetic susceptibility, yielding this implication:

If D(N)=P(N), then A(N)=A(P).

Zachary Voch said...

Anne Buchanen: "But I think we all basically agree, and are essentially just fiddling with the issue of why we're looking for cause anyway."

I think our purposes in determining causal chains are precisely the same, prevention/treatment of illness being the goal so far as I can gather.

I think that we agree that mechanistic explanations are often over-simplified. I think that we (in this case I mostly mean Dr. Weiss and myself) disagree over whether or not modern research science, particularly the area of genetics and focus on molecular explanations, suffers from a conceptual inability to address nonmolecular causes in explaining and preventing illness due to solitary focus on the molecular `proximate' causes. Perhaps it is due to the particular examples we discussed so far, but these situations seem to mostly center around nonmolecular causes (or correlates) that are already widely known and well established (hence not items for research while also not ignored) and less understood or as of yet untreated problems on the molecular level which demand this sort of attention, all while preserving the same purposeful approach in fighting disease. Which brings us to this:

Ken Weiss: "I guess I come down to what the purpose of your search for cause is anyway..."

I think here we agree in addressing the broader network of causes, but we (might) disagree where I think that specialization resolves this issue. I do not think that biochemists should, by virtue of their degree and chosen careers, split their focus between 'HIV vaccine research' and `fight to eliminate proverty.' As far as research has already established, both are worthy `causes' (couldn't resist). Like the HIV/AIDS hypothesis, the obesity/diabetes theory is not at the `research problem' level by any reasonable measure; it's at the `public awareness' and `social impact study' level. The open science that remains here, with the purpose being to eliminate Type II diabetes/AIDS, is research matter for sociologists and psychologists. With the example of Collins and the NIH the mind, we can ask whether or not expenditure and public outreach are disproportionate to their impact as compared with molecular research, but public outreach isn't ignored by the NIH. Our complaints might be about emphasis and resource management, not ill-considered causal
chains.

I would love to explore the problem of causal simplification at the level where we certainly agree it is a frequent problem (and I would say partially out of necessity, but that's based on experience with nongenetic examples), at the mechanistic level. I'll take a look at the Waters paper (and those posts) once finals week finishes up. It's been a while since I've had the time to lurk regularly.

Thanks for keeping a layman entertained.

Anne Buchanan said...

Two points, one minor, one less minor.

You say: "That is, if dietary factors in developing type II diabetes are controlled, we expect a higher incidence of the disease in Native American populations than the general population. We would say the case of this implication alone is genetic susceptibility."

I didn't explain that very well. In fact, the idea is that there's a gene by environment interaction here such that without the provocative environment, Native Americans are free of diabetes (the last 20,000 years), and without the susceptibility genes, even given a provocative environment, risk is also less. (Of course, most populations are susceptible to t2d on a 'western' diet, but the natural history of the disease is somewhat different in different populations, suggesting varied genetic responses.)

A larger quibble I still have is with your argument that oversimplified causation isn't a problem. The genetics of the diabetes epidemic have been studied for 40 years -- that is literally searches for the gene or genes for type 2 diabetes. This is clearly neither the explanation for the epidemic nor the cure.

And of course it's not just diabetes where this is the common approach. Asthma, autism, schizophrenia, heart disease, cancers, etc etc have all been studied to death from the same perspective. You can find the same reductionist approach to normal traits like behaviors, intelligence, risk taking and so on.

The larger epistemological issue is whether these kinds of complex traits can be understood with the scientific methods we have, which are so good at reductionism. As Ken said, "The drive to find single or simple causal explanations, when there is reason that we should be more comprehensive, is where we see problems."

(Sorry about the delayed responses -- didn't see your comments until today.)

Zachary Voch said...

@Anne Buchanan

"I didn't explain that very well. In fact, the idea is that there's a gene by environment interaction here such that without the provocative environment, Native Americans are free of diabetes (the last 20,000 years), and without the susceptibility genes, even given a provocative environment, risk is also less."

I should have been clearer in my reply as well. It was based on this point that I assigned primacy to the environmental factor over the genetic susceptibility. Up to type 2 diabetes, the causation of the gene is dependent on the causation of the diet. Else, we're dealing with type 1 diabetes; in this case, genetics and early developmental factors gain primacy over diet.

The need of the `provocative diet' for the higher incidence of type 2 diabetes in Native Americans was implicit. It is why I described the action of NA genetics on incidence of diabetes via implication

If D(P)=D(N) then A(N)>A(P).

I think that your point is that if D(P)=D(N)=`recommended/healthy', that A(N)=A(P). What I left unstated is the assumption that D(P) was assumed to be `unhealthy.' It was also why I said that lowering D(P) was more important (up to reduction of type 2 diabetes) than correcting the inequality.

But, as far as actively fighting against diabetes, my original point holds that, up to research, the implication `D(P) increases (resp. decreases) with A(P)' is well-established. This brings us to the `large quibble':

"A larger quibble I still have is with your argument that oversimplified causation isn't a problem. The genetics of the diabetes epidemic have been studied for 40 years -- that is literally searches for the gene or genes for type 2 diabetes. This is clearly neither the explanation for the epidemic nor the cure."

I have yet to hear a doctor or scientist say that genes alone are responsible for the diabetes epidemic to the exclusion of diet, and it would greatly surprise (and disappoint) me to hear otherwise (without some amazingly strong and counter-intuitive research on hand).

Is it a problem that genetic factors have been studied so extensively? If diet were ignored as a component, I would say yes, but I think the problem here is that the relationship of diet and type 2 diabetes is a closed problem: everybody knows it. The open problems are things like `why are Native Americans more likely to develop type 2 diabetes?', problems that appear to have a genetic answer. There are plenty of other open problems concerned with genetic risk factors for type 2 in the more general population.

But, given diet as the primary cause and deciding factor where genes play a secondary/complementary role, the major responsibility for fighting the epidemic lies with the public: this issue isn't one of a poor research paradigm or simplified model of causation.

The minor responsibility lies with researchers on the molecular level.

"The larger epistemological issue is whether these kinds of complex traits can be understood with the scientific methods we have, which are so good at reductionism."

Just curious, but what scientific method might be an improvement?

Ken Weiss said...

Zach,
You raise too many points to go over with the care they deserve. They represent a struggle to come to grips with the causal landscape and the way that different views, and different approaches can exist or coexist. But there are too many points for the time I have at the moment, and I don't want to give superficial answers (that I might regret!). It's a busy time here, and I hope I can respond in the future.

Our hope is to make this blog food for thought, and this is just the kind of exchange we like.

Meanwhile, if you have finals to take, you had better put your thoughts to them!

Anne Buchanan said...

A few quick thoughts (which, as Ken says, don't do your comments justice, but this will have to do for the moment).

We do need to keep in mind that that what one considers a cause largely depends on why one's asking. Is it for basic or applied scientific purposes?

It my be that doctors all do believe that causation is complex, but since they are primarily thinking about treatment rather than prevention (more on that in a bit), they are free to focus on diet and exercise. And in fact geneticists or epidemiologists may _say_ they understand that causation is complex, but when they actually design their studies, it's primarily reductionist because that's what they know how to do and what our methods are good for.

But now that we can all send cheek swabs to companies like 23andMe or DeCodeMe to be told what diseases we are at genetic risk of, a doctor might in fact be faced with a patient who comes in with a printout that tells him/her s/he's at high risk of t2d. What does the doctor do with a statistic like 25, 30, 60% risk of diabetes? In many cases, s/he will treat it as if it's 100% risk. So, de facto single causation. Of course this won't always be true with every disease, and it depends on the disease and on what the preventive steps are, but consumers are certainly assuming genetic causation when they send their money to 23andMe. And how can a doctor explain, or even understand, the kind of risk estimates these companies provide -- when each study comes up with different estimates, and they keep changing?

Finally, if I knew what kind of methods would solve complexity, I'd be rich and famous and off on a tropical island somewhere! Many people smarter than I understand that our methods aren't really working -- geneticists are looking at gene pathways, biologists at biological systems, epidemiologists at interacting social and biological causative factors, and so on.

But, why don't we know what causes asthma? Or why not everyone who smokes gets lung cancer?

Anne Buchanan said...

A good example of how poverty 'causes' AIDS. http://www.nytimes.com/2010/05/10/world/africa/10aids.html?hp

...all the nuances and complexities you bring up notwithstanding, Zachary Voch!