[Help] [Aide] [Up]

The Cancer Journal - Volume 10, Number 3 (May-June 1997)


The structure of the disease "cancer"

The results just published by Bailar and Gornik (1) support the initial conclusions by Bailar and Smith (2), 11 years later. In the US, age-adjusted mortality due to cancer continued to rise in adults until 1994 and, since then, has fallen very slightly, but in men only. The fall cannot be attributed to improvements in treatments and has little to do with earlier diagnosis of the disease ,but is related to a decrease in cancers due to smoking.

Are hopes of genuine advances over today's treatments therefore just wishful thinking ? Will we have to be satisfied with preventive measures as Bailar suggests in both papers ? We hope not. Although we agree that more effort should be placed into research in prevention, we consider that substantial progress could also be made in therapy if we were only willing to study the structure of the disease. We shall attempt to explain what we mean by this expression.

Since the beginning of the 19th century, diseases have been studied by an anatomo-clinical approach which has been strengthened by the availibility of biochemical and technical tools, in particular of imagery and computing. However, although we are now able to obtain the image of a lesion or even molecular biology data, the old reasoning of the 19th century still prevails : there must be a direct relationship between these structures and their function, and structural abnormalities are reflected in the elements that make up the pathology. In cancer research, this logic has meant that cancers are classified on the basis of criteria relating to anatomy and histology (TNM or AJC classification). Such a classification does define relatively homogeneous categories when the treatment is either surgery or radiotherapy but is far less meaningful when drugs are prescribed (chemotherapy, immunotherapy, hormone therapy). The anatomically defined groups are heterogeneous with respect to these treatments, even if we know that this organ or that histological tumour type is statistically more of less likely to respond to first-line radiotherapy or chemotherapy. By opting for a single classification, we have locked ourselves in a one-dimensional world*.

What applies to clinicians, also applies biologists. It is partly for this reason that neither high-performance clinical/radiological/biological tools nor the reams of biological data obtained with great technical expertise have led to any significant decrease in mortality rates in adults. The increasinglytechnical nature of medicine and the requisite and unavoidable need for reductionism generate a wealth of valid data but no-one seems to know how to translate this into progress.

Statements such as 'diseases are complex phenomena', 'complexity cannot be reduced', 'the whole is greater than the sum of the parts', and 'we must perform multidisciplinary or transdisciplinary clinical research' have become platitudes. They may have some relevance but if the one-dimensional attitude does not change, laboratories will keep on testing the same working hypotheses and clinicians will continue to be torn between unrealistic hopes, on the one hand, the hope of a specific, effective miracle drug found by computer-aided design and molecular biology techniques and, on the other, the hope of an unexpected innovation born through serendipidity and tenacity. There is no similarity between reductionism and a one-dimensional approach. Reductionism is too often stigmatised as a major fault in experimental research and, to our mind, this is an error of judgement. Reduction is virtually unavoidable when we perform a fine analysis of complex phenomena. The error is in not proceeding further and in believing that the analysis is finished once the study has been dissected in a piecemeal fashion. However, these pieces produced by a reductionist study must now be used to rebuild the structure according to a procedure classically known as a synthesis.

By stepping out of a one-dimensional into a multidimensional world, we can begin to study the structure of the disease 'cancer'. Because it is not possible to characterize all the variables that define this structure, we have to do as with any other complex object and examine it from as many sides as possible. For this, at least three conditions must be met : first, we must be convinced of the absolute need for a multidimensional approach to step out of the quagmire; second, we must not abandon existing variations on the anatomo-histological classification insofar as they are useful but must introduce new classifications on the basis of other variables known to be relevant (examples are given below); third, we must adopt descriptive methods of data analysis that, unlike classical statistics (3), unravel, step-by-step, the structure of a multidimensional system and thereby elicit new scientific hypotheses.

This objective descriptive approach implies that even the most widely accepted truths have to be questioned. For example,
- are all cancers genetic diseases ?
- do they all evolve linearily from a single cell ?
- does the death of cancer patients occur by invasion of vital organs by neoplastic tissue from the primary tumour and/or from metastases ?
etc..... etc...... Each person will thus question the truths most meaningful to him or her and reflect on the many possible answers. If we accept that there may be more than one answer, dogma itself - and the institutions in which this dogma prevails - are brought into question. Although this salutary little exercise is a prerequisite to understanding the multidimensional structure of diseases, it is not sufficient.

What is also required are new classifications of cancers. These can be obtained by algorithms that automatically partition populations of patients on the basis of preselected variables. Or, inversely, one can predefine partitions and then try to discover which sets of variables are discriminatory. Both methods can be the basis of new working hypotheses. For example, why not separate :
- all cancers, regardless of site and histology, that are resistant and not sensitive to first-line chemotherapy. This partition might help identify further factors important in drug resistance,
- all occult cancers that do not progress versus all cancers of the same size that do.
According to his or her interests, each research-worker and clinician will make his/her own classifications.

The scope of research is restricted by overexploited mechanism-based hypotheses. A large number of laboratories employing a large work-force devote themselves to few research topics whilst spending most of the financial ressources. And they continue to do so, without hesitation, even in the absence of conclusive results that help improve therapy. There are better ways of doing things in our effort to stabilise and cure more patients, and thus decrease mortality rates. The conclusions of Bailar and Gorik's study suggest that the time is ripe for a rigorous appraisal of concepts and hypotheses. Living huddled together in an oasis may be comfortable and reassuring but leaving protected areas to irrigate the desert is more worthwhile.

Jean-Claude Salomon
To contact the author...Click here Thank you.

1. Bailar JC III, Gornik HL. Cancer undefeated. New Engl J Med 336,1569-1574., 1997.
2. Bailar JC III, Smith EM. Progress against cancer ? New Engl J Med 314, 1226-1232, 1986.
3. Bansard J-Y, Kerbaol M, Salomon J-C. Some thoughts on correspondence factor analysis. Cancer J. 9, 110-112, 1996
4. Salomon JC. Cancer, tumors, and paraneoplastic syndromes. In: Iversen OH, ed. New Frontiers in Cancer Causation. Proc. 2nd Intl Conf Theories of Carcinogenesis. Washington: Taylor and Francis. 1993, 73-80.

*"Biology is divided between two cultures. One is dominated by the theories and concepts of cellular and molecular biology, based on the accumulation of simple scientific results. These results are irrefutable, numerous and one could almost say final, and they concern the specificity of structure : function relationships in the phenomena under study. Some of the most obvious examples are the increasingly fine description of the mechanisms for passing genetic information from one generation to another (or from the center of a cell to the periphery), a vision of the regulation of these mechanisms improving daily, and the nature of the relationship between a receptor and its ligand, a model which now seems almost universally applicable.

The other culture, sometimes called phenomenological, affirms, often in the defensive manner associated with a truism, that the whole is more than the sum of the parts and that a mass of facts cannot replace a scientific understanding of physiological and pathological processes taken in their full complexity. This position, irrefutable and incantatory, is not making any greater contribution to medical progress at the moment.

At the moment we find ourselves in a crisis which should be examined by epistemologists, historians, scientists and doctors. We have a surfeit of techniques and of data and a lack of theories that are able to link the different levels ranging from the molecular to that of the human being or of the population." (4)