Georg Ivanovas From Autism to Humanism - systems theory in medicine
2.1 The medical paradigm
|Classical science in its diverse disciplines, be it chemistry, biology, psychology or the social sciences, tried to isolate the elements of the observed universe chemical compounds and enzymes, cells, elementary sensations, freely competing individuals, what not expecting that, by putting them together again, conceptually or experimentally, the whole or the system cell, mind, society would result and be intelligible.” (Bertalanffy 1968: xix)|
The current medical paradigm is characterized by two main features: First, modern medicine is based on measurement. By that observations are quantified and experiments become comparable. Second, in order to cope with the complexity of human physiology and pathology the observed phenomena are broken down into defined parts, a procedure called reductionism.
Under the conditions of such an experimental setting a certain range of observation is defined. In this range parameters and their changes through interventions can be coherently tracked. Thus, the concept of cause and effect is introduced.
The received descriptions and reactional patterns (physiology, pathology, biochemistry, etc) are called specific. Current medicine speaks of ‘specific’ diseases like rheumatoid arthritis, mononucleosis or diabetes mellitus. Interventions to alter such specific pathologies are called ‘specific treatments’. Methotrexate in rheumatoid arthritis, insulin in diabetes or neurotransmitters in depression represent such treatments.
This is, in short, the foundation of modern medicine which dates back to the middle of the 19th century. Most important in its development was Claude Bernard (1813-1879). He was not only an outstanding observer and scientist. He also formulated the basics of modern medicine. He anticipated as we will see biological and medical models that arose only lately. His book An Introduction to the Study of Experimental Medicine (1865) is, in fact, a masterpiece of medical thinking and it served as an introduction for medical students still a hundred years later. Most of his statements are valid until today.
However, Bernard was a contemporary of Laplace and he strongly believed in determinism which turned out to be wrong. He wrote: “Confidence in absolute determinism in the phenomena of life leads….to real science” (Bernard: 69). “When once the conditions of a phenomenon are known and fulfilled, the phenomenon must always and necessarily be reproduced at the will of the experimenter” (Bernard: 67-68). “But the real and effective cause of disease must be constant and determined, that is unique, anything else would be a denial of science in medicine” (Bernard: 83). “Absolute determinism exists indeed in every vital phenomenon; hence science exists also” (Bernard: 65).
He also formulated already the principles of reductionism: “When faced by complex questions, physiologists and physicians, as well as physicists and chemists, should divide the total problem into simpler and more and more clearly defined partial problems” (Bernard: 72).
Determinism and reductionism together constitute the machine model, the paradigm of Bernard’s time: “A living organism is nothing but a wonderful machine” (Bernard: 63).
A medicine based on such a model does not differ fundamentally from the work in a garage where sick humans are handled like a broken car (Ahn et al 2006a). This idea has not changed since and some even expect no change for the future. “If you go to your surgery twenty years from now complaining of rheumatism, your doctor may well check out the relevant section of your personal genome CD-ROM rather than reach straight for the prescription pad.” (Day 1998a).
New Scientist provided a collection of similar opinions: “Malcolm Lader, a psychiatrist at the Institute of Psychiatry in London: 'I don't see any fundamental technical obstacle to altering personality with drugs. After all, the traits that make up personality are rooted in neurochemicals.' And from Jerome Kagan of Harvard University, who studies the biological basis of shyness: 'Fifty years from now we may have drugs that can alter personality profiles. Things are moving very fast'“ and creating the following scenario “Feeling irritable and melancholic, you reach for your computer and call up Normopsych, an on-line drugs service specialising in personality restructuring. After downloading your life history and personality profile data and completing virtual reality tests of rejection sensitivity and mood, you sit back in your chair. A few seconds later the screen fills with a rotating, three-dimensional image of the brain. A handful of neurotransmitter pathways are flashing ominously. The diagnosis reads: 'Serotonin levels 15 per cent below par in limbic system. Boost with 100 milligrams per day of MoodStim and AntiGrief.' “ (Concar 1994).
Although these are somehow exaggerated opinions, they express the common understanding of medicine nutured by the expectation that “although the road ahead is long and winding, it leads to a future where biology and medicine are transformed into precise engineering” (Kitano 2002).
But there is, as we all know, a problem with this trivialization of living processes. This shall be demonstrated with the ‘polypill’. The polypill is a mixture of eight different drugs and vitamins effectively used in contemporary medicine. The pill should be provided for every patient over 55 with the expectation to reduce mortality of cardiovascular disease for about 80% (Wald/Law 2003). Astonishingly, most of the medical scientific community took this paper seriously and discussed the pros and cons at length. It is still held as an option worth of consideration (Smith 2005a). The example of the polypill could be refuted simply by the fact that it violates Simpson’s paradox (chap. 2.5.d). But this does not satisfy the basic epistemological question whether reductionist findings might somehow be synthesized and add up to constitute a whole.
Simple additivity surely is no solution, as additivity has the following pre-equisites: „The first is that interactions between ‚parts’ be nonexistent or weak enough to be neglected for certain research purpose. Only under this condition, can the parts be ‚worked out,’ actually, logically, and mathematically, and then be ‘put together.’ The second condition is that the relations describing the behaviour of parts be linear; only then is the condition of sumativity given, i.e., an equation describing the behaviour of the total is of the same form as the equations describing the behaviour of the parts; partial processes can be superimposed to obtain the total process, etc.” (Bertalanffy 1968: 19).
The belief in a simple additivity as expected with the polyplill is an exception and even Bernard never had such a trivial sight: “Moreover, as we know, it happens that properties, which appear and disappear in synthesis and analysis, cannot be considered as simple addition or pure subtraction of properties of the constituent body” (Bernard: 90).
But how is it possible to synthesize the different findings provided by reductionist research? A solution often tried today is computer simulation. “The challenge …is to develop mechanistic models that begin from what is understood (or hypothesized) about interactions of the individual units, and to use computation and analysis to explain emergent behavior in terms of the statistical mechanics of ensembles of such units” (Levine et al 1997). Probably every physiological mechanism has been simulated, including the heart (Noble 2002) and the brain (Graham-Rowe 2007). Such models are used to investigate in silicone (and no longer in vitro or in vivo) details of physiology, the impact of medications, the influence of environmental changes, etc.
But is computing the solution of the question whether reductionist research put together create a whole? Surely not, therefore “the result is that these models produce cartoons that may look like nature but represent no real systems… Yet it is fair and important to ask how seriously such predictions should be taken” (Levine et al 1997).
First, there is a problem that seems to be technical. If a model is too simple, it neglects essential mechanisms of the real system, limiting its potential to provide understanding. If it uses too many facts the simulation becomes too complex and it will get lost in details (Grimm et al. 2005). Thus, complexity management is caught between the Scylla of reductionism and the Charybdis of arbitrariness (chap. 6.13). The actual problem, it seems, is not the lack of biological data or of computers. The main problem is the lack of a proper understanding of biological processes.
A central question in all these attempts is whether models are constructed ‘bottom-up’, i.e. from cellular compartments to the whole or ‘top-down’, from the whole to the cellular compartments (Noble 2002). Due to the reductionist foundations of medical science, the bottom-up approach is mainly favoured. Gene function and molecular action is computed in the hope to create reliable models (Mood et al 2004). Implicit is the expectation that the „knowledge of the genetic architecture will lead to increasingly realistic models of social evolution, while identification of the products of major genes can elucidate the molecular basis of social behavior “ (Krieger/Ross 2002). The same happens in brain research: “The truth is that we need to grasp what's happening at the cellular and molecular level before we can begin to tackle the essence of brain function” (Wallis 2000).
However, what is not answered and what is even not understood as a problem is the basic epistemological question about the relation between the whole and its parts. How can there be an ‘upward causation’ from the behaviour of the parts to the behaviour of the whole system. The central issues of logical typing (chap. 3.2) and of emergence (chap. 4.10) are normally neither addressed nor understood.
Also the ‘top-down’ model, favoured in this thesis, has to account for how a ‘downward causation’ (how the whole might influence the parts) can be possible (chap. 4.8).
The knowledge of a functional circle (may it even be as complex as a simulated heart), does not imply knowledge about the whole system. The reaction of ‘wholes’ cannot be predicted by that (Bertalanffy 1968: 149-153). There is fundamental unpredictability (von Foerster/Bröcker 2002: 175-179) due to the circular (cybernetic) function of parts (chap 4.4). The behaviour of systems is not necessarily a consequence of the behaviour of parts.
Another limitation to computer simulations is the concept of an environment. The environment of a simulation is the knowledge and the theory of the scientist. Thus, computer simulation is still a reductionist method which only works a little faster. But it creates no real difference to the usual experimental setting.
Reductionist research excludes unpredictability by limiting the number of factors. As in reality factors are numerous, we have an artificial situation, true if and only if the factors behave in the way they are simulated, i.e., under stable conditions. Thus, the results are only true under the defined circumstances.
This shall be demonstrated with the following example: A friend of mine investigated the turnover of chylomicrones and LDL in rats. In the biological system he found a deviation of about +/ - 2000 %. When he perfused the isolated liver the deviation between the experiments fell to about +/- 100 %. Working only with prepared hepatocytes he reduced the deviation to about +/- 25-50 % and with the standardized HEPATOMA cell-line to +/- 5-10 %. And even this deviation vanished when he was able to clone a liver cell. Then his results became identical (Retzek, private communication).
In normal life unexpected events might totally change the outcome of a process. For example, decision making is a complex procedure. It has been shown that short term effects and long term effects are cautiously considered and then a decision is made. Brain scanning indicates that the ventral and dorsal striatum is involved in this process (O’Doherty 2004). It also could be demonstrated that men and women tend to prefer long term benefits. However, if during the process of testing the men see a photography of a beautiful girl they throw all long term considerations over board and look only for a short term benefit, something not true for women if presented a photography of beautiful men (Wilson/Daly 2003). This is supported by another experiment that demonstrated that in short term decisions mainly parts of the limbic system associated with the midbrain dopamine system, including paralimbic cortex are involved. These centres are associated with the emotional aspects in decision making (McClure et al 2004).
These experiments shows that if only a small detail is changed in an otherwise stable setting the outcome differs fundamentally. It is probably known since humans exist that as soon as emotions come into play the results become in a way unpredictable. Somehow they spoil everything. In the daily practice all kind of emotions are involved which do not match the experimental setting. There are always unexpected developments. Such developments are normally called psychological or psychosomatic. But this is not correct. Unexpected behaviour arises all over and no computer simulation is able to foresee that.
This difference between reductionist setting and real life might explain the following case: The Randomised Aldactone Evaluation Study (RALES) demonstrated that spironolactone significantly improves outcomes in patients with severe heart failure. The publication of RALES was associated with abrupt increases in the rate of prescriptions for spironolactone. This had no beneficial effect, but lead to an increased hyperkaliaemia-associated morbidity and mortality. (Juurlink et al 2004).
The authors of the control study and of the related editorial (McMurray/O'Meara 2004) assumed that physicians had made mistakes (not observing risk factors properly, not controlling potassium levels often enough, etc). But may be this view is just too simplistic. In everyday life there might have been factors which just do not exist in the somehow reductionist setting of the hospital. The therapy might have increased rigidity (chap. 6.4). It also might have altered the ‘information’ of the drug (chap. 4.1). In such a case a rare event might lead to the breakdown of the whole system.
These reflections shall not serve as an exhaustive explanation of the difference between the reductionist setting and real life. Though, they might give an impression about the difficulties translating experimental results into practice. The fact as such is well known. But scientists often do not have the appropriate epistemology to explain the differences and the difficulties they encounter.