To return to the home page, click here
Notes for a lecture
ON CAUSALITY AND THE PROBLEM OF ANEURYSMS
M. David Tilson
Ailsa Mellon Bruce Professor of Surgery
Columbia University and the St. Luke's/Roosevelt Hospital
1000 Tenth Avenue, New York, New York, 10019.
Phone: 212-523-7779; Fax: 212-523-6495
Internet address: mdt1@columbia.edu
-----------------------------------------------------------------------
I believe that the term "atherosclerotic aneurysm" has been very
unfortunate, because it has given us the notion that we have identified
the necessary cause, the same as we have in the case of "syphlitic"
aneurysm. In what follows, I will make the argument that atherosclerosis
does not provide an informative explanation of aneurysm disease.
Arguably the most incisive contemporary philosopher on the subject
of causality is Paul Humphreys at the University of Virginia
(Charlottesville). He presupposes an ontology where things are presumed
to be real and where causal relationships exist among things that are
discoverable by experimental methods. This ontically-oriented metaphysics
enables him to avoid the difficulties of Hume, who was unable to discriminate
causal from non-causal relationships among observable regularities. I will
return to this idea later when we consider joint observable effects at the
root of the fallacy of spurious causation. And, eschewing linguistic models
that articulate rather sterile notions of causality (in my opinion),
Humphreys uses the concept of probability to avoid the problems inherent in
formulations that fail to deal adequately with the situation of chance
occurrences.
Reduced to simplest terms, his view is that we can properly assert
that X causes Y if the existence of X increases the probability of the
existence of Y. This is a useful notion. The classical concepts of necessity
and sufficiency are the boundary conditions, where the probability factor is
either zero or 100%. X is a sufficient cause of Y, if the existence of X
makes Y inevitable. X is a necessary cause of Y, if in the absence of X,
Y will never occur. I will argue in what follows that zero and 100% are
imaginary numbers in science.
THE ANEURYSMS OF MICE AND TURKEYS
Let us see how these ideas work out when applied to the causation of
various forms of aneurysm disease. Among the most intuitive cases under
discussion today for a necessary and sufficient cause is mutation of a
recently discovered gene for a metal-transporting-ATPase at the mottled
locus of the Brindled mouse, which is closely related to the Blotchy. The
Blotchy mutation is sufficient in the sense that 100% of the males carrying
the mutation will develop aneurysms, if they do not die of other causes
before six months of age. It is necessary in the sense that the phenotype
does not occur in the absence of the mutation. You get a perfectly normal
C57BL6 black laboratory mouse.
However, sufficiency fails under the condition of a negative
contributing cause. We have found that propranolol reduces the probability
to 70 % at six months. Nor can we be sure that only one gene will satisfy
the requirement for necessity, since one cannot a priori rule out the
possibility that a new mutation may arise that causes identical phenotypic
effects. Since we cannot be sure that a cause is absolutely necessary or
absolutely sufficient until the world is over, I will henceforth use the
terms in a less formal sense, when I intend to imply that their presence or
absence makes the existence of the respective effect highly probable.
The example of the broad-breasted-white (BBW) turkey allows further
exposition of probabilistic causality and enables us to examine an example
that has multiple causal influences. The broad-breasted-bronze (BBB) turkey
does not get spontaneous aortic aneurysms, so we may infer that some feature
of the genome of the BBW-turkey is a necessary cause, in the sense that if
this feature were not present, the aneurysms would not occur. However, the
genetic factor is not a sufficient cause, since some BBW-turkeys do not
develop aneurysms. The case of multiple causal influences applies because
there are well-defined positive and negative contributing causes. If the
BBW-turkey is treated with the lathyrogen, beta-amino-proprionitrile (BAPN),
the probability of aneurysm development increases substantially. If the
BBW-BAPN turkey is treated with the negative contributing cause, propranolol,
the probability is reduced.The example of the BBW-turkey also allows us to
introduce the notion of a chain of causal events. What is the intermediate
step between administration of propranolol and a reduced probability of
aneurysms? The hypothesis that propranolol achieves its status as a negative
contributing cause through its antihypertensive properties is defeated by
the fact that other antihypertensives are not effective, even though a
similar hemodynamic response is achieved. It now appears that propranolol
is a negative contributing cause, both in the BBW-turkey and in the Blotchy
mouse, because of an effect that stimulates crosslinking in connective
tissue.6
NON-HUMAN PRIMATES
Now consider the examples of the non-human primates. As in the
BBW-turkey, genetic considerations apply. Aneurysms are thirteen times
more common among cynomolgus monkeys than rhesus monkeys. Among cynomolgus
monkeys that develop aneurysms, it would be extremely interesting to know
if they comprise a kinship, of brothers, fathers, and/or monkeys' uncles.
Zarins has reported aortic ectasia in four of six monkeys subjected
to a regimen of atherogenesis and regression. The monkeys were fed a 2%
cholesterol / 25% peanut oil diet for 6 months, and their mean serum
cholesterol reached 788 mg%. They were then fed a 0.25% cholesterol /
15% corn oil diet for 6 months. There was less than a 50% increase in the
mean aortic radius, by comparison to the 6 month atherogenic diet controls.
The authors claim that one animal had a "manifest" aneurysm on the basis of
a three-fold increase in lumen area. From data provided in the paper, one
can calculate that this animal had an increase of approximately 65% in
aortic diameter. Using one conventional criterion of a 50% increase in
diameter to define an aneurysm, this animal makes the cut, but not by much.
Controls were not reported for the regression diet alone. The
information at hand does not rule out the possibility that the regression
diet caused an aneurysm in one monkey. Suppose that there had been an
appropriate control group. Would one positive monkey in six be statistically
different than no positive monkeys in six? By Fishers Exact Test of a 2 x
2 contingency table, the one-tailed probability value = 0.50. It would take
about 20 animals in each group to avoid a Type I error.
Was hypercholesterolemia a contributing cause in this experiment?
Perhaps, reserving judgement until there is more than one monkey on which to
base an opinion and until there are additional controls.
THE HUMAN CONDITION
Finally, we come to the species of greatest interest. A mutation
in the gene for fibrillin on chromosome 15 (Fib-15) is a necessary and
sufficient cause for Marfan's Syndrome. Certain mutations in the gene make
the phenotype inevitable. Also, so far, no case of the Marfan phenotype
has been shown to have two normal genes for Fib-15. The case of the
abdominal aortic aneurysm (AAA) could turn out to be as decisive, if a
similarly definitive single major locus were discovered to be the necessary
cause. It would then be left to determine whether the gene satisfies the
condition for sufficiency, if an affected individual lives long enough to
pass through the window of peak risk. My hunch is that things may not be so
simple, since the AAA may be a disease of greater clinical and molecular
heterogeneity.
Vascular surgeons from time to time see patients with ordinary
non-specific AAA's, who have normal cholesterols and minimal systemic
atherosclerosis. Thus, atherosclerosis is not a necessary cause for
aneurysm disease in man. Is it a sufficient cause? Absolutely not.
From the point of view of risk factor analysis in the AAA, it
appears that smoking is a contributing cause, insofar as the presence of
the risk factor increases the probability of the disease. But smoking is
neither necessary nor sufficient. Smoking is also a contributing cause for
atherosclerosis. The aneurysmal aorta has features that are different from
the atherosclerotic occlusive aorta, specifically adventitial inflammation
and weakening of the structural integrity of adventitial collagen, which
permits ballooning. Atherosclerosis, on the other hand, is primarily a
disease of the subendothelium, associated with the migration and
proliferation of cells that may lead to stenosis of the lumen.
Consider the hypothesis that the mechanisms by which smoking
increases the probability of both diseases are different. Smoking inhibits
the activity of alpha-1-antitrypsin, which may contribute a positive causal
influence for aneurysms, by interference with the mechanisms of defense
against enzymatic destruction of connective tissue. Smoking may also result
in the exposure of subintimal smooth muscle cells to mitogens absorbed from
burning tobacco, which may contribute a positive causal influence for
atherosclerosis. If a person has a genetic susceptibility for both diseases,
one would predict that both aneurysmal dilatation and atherosclerosis will
develop. This analysis can be extended by analogy to overlapping Venn
diagrams in the field of a common contributing cause.
Correlation of joint effects of a common cause can lead to the
fallacy of spurious causation.2 To borrow an analogy from Humphreys, we may
observe that when the water in an old-fashioned barometer spout rises, a
storm often follows. To say that rising water in the device causes the
storm is the fallacy of spurious causation. However, it is informative
scientifically to say that falling barometric pressure has the joint
observable effects of water rising in the spout and rain falling. A similar
analysis applies not only to smoking but also to possible joint effects of
other factors like hypertension and hypercholesterolemia in aneurysm disease.
The occurence of joint effects may permit accurate predictions, but joint
effects do not provide informative causal explanations.2
THE ATHEROSCLEROTIC DOGMA
Pyeritz has observed that Marfan described the syndrome that bears
his name in the late 19th century, but it was not until about 1950 that
skeptics were convinced that there was a necessary cause (in the sense of a
single major locus). Appreciation of the need to postulate a necessary
cause for the AAA was even more delayed, in large part (I believe) because
of the unproven dogma that the disease is caused by atherosclerosis. One of
my students, Anil Hingorani, has searched for the first historical
attribution of the disease to atherosclerosis, and he found that it is
traceable to the translation into English of a book by the Italian surgeon,
Scarpa, in the early 19th century. Although the notion of a difference in
atherosclerotic stenosing versus atherosclerotic dilating processes had been
raised before, I believe that my paper in 1980 was the first systematic
challenge to the idea that atherosclerosis is causal. Once the ground was
broken, seeds quickly sprouted in many laboratories in the United States and
around the world. It is astonishing how long the disease had been neglected,
simply because it was so widely presumed that atherosclerosis was an
informative explanation. The old dogma has been passionately defended by
the faithful, sometimes with religious fervor, against attacks by the
blasphemers; but that has not stemmed a dramatic crescendo of new knowledge
on the subject.
CHAIN OF CAUSATION IN THE AAA
As Humphreys observes, scientific explanations are progressively
discovered. And, in complex situations, the matter of timing can lead to
confusion. He makes the following argument to illustrate a case of
sequential overdetermination. If I light a long fuse leading to a bomb, and
you then light a much shorter fuse, which ignites the bomb first, it was you
and your fuse that caused the explosion, not me and mine. "One must work
backward in the completed process that leads from cause to effect in order
to arrive at the correct attribution." 2
Working backwards, I have advocated the notion that mechanical
failure of the collagen is very proximate to the aneurysm effect. Since
the time that I wrote a chapter with a subsection entitled "Why collagen
must fail in aneurysm disease," this idea has gained support. The evidence
for increased collagen turnover and collagenolysis is impressive. We have
shown the presence of collagenase in soluble extracts of aneurysmal aorta
and also shown that it is detectable immunohistochemically in mesenchymal
cells and occasionally endothelial cells. Gelatinase-B (also known as
Matrix MetalloProteinase 9 or Type IV Collagenase) and stromelysin-1 are
also present and localize to macrophages.25 In addition, plasmin and the
plasminogen activators are present.
The complex interrelationships of the contributing causes, including
the several types of cells present, the immunoglobulin response, the
cytokinetic signals, and the biochemistry of the enzymes, will all be
discovered in due course by the scientific method. And, so will the
necessary cause, which I still hypothesize to have a genetic basis.
IN CONCLUSION
The most extreme iconoclastic position would be that the present
evidence supports a stronger argument that aneurysms cause atherosclerosis
than the other way around. Consider the subclavian artery aneurysm that
begins as a post-stenotic dilatation caused by a cervical rib. In due
course it will develop atherosclerosis. Rob once reported calcification
of a post-stenotic dilatation of the aorta distal to a coarctation.
Bauer Sumpio and I found that there are boundary layer separations,
turbulence, and reversal of flow on the surface of a glass model aneurysm
of the human aorta. These conditions at the flow surface have been shown
by many workers to stimulate the development of atherosclerotic lesions.
It should not be surprising that Reed and coworkers found a large portion
of the flow surface of an aneurysm to be affected by atherosclerosis.13
I remain skeptical that primary attribution of aneurysm disease
to atherosclerosis has been particularly helpful in the search for a
necessary cause, and I feel that I have been very fortunate to see some of
my ideas serve as a propaedeutic that has brought many talented young
investigators into the aneurysm research field. A natural scientist cannot
rest until there is an informative explanation for a biological event.