Non-dits méthodologiques en recherche médicale Methodological unsaid in medical research

Résumé

La pratique médicale moderne doit s’appuyer sur une « médecine fondée sur des faits prouvés » : mais qu’est ce qu’une preuve en médecine ? Pour certains philosophes, pour qu’il y ait preuve, il faut qu’il y ait croyance : quelles conséquences cela peut-il avoir dans le domaine médical ? Pourquoi doit-on tirer au sort l’allocation des traitements dans un essai thérapeutique ? Pour rendre les groupes semblables ? Mais pourtant, en cas de malchance, le déséquilibre peut être important… Cet article aborde plusieurs notions souvent utilisées par les méthodologistes : causalité, preuve, croyance, aléa, tirage au sort, modèle, observation, expérimentation, critère d’inclusion, mesure. L’objectif est de sortir du discours habituel, parfois un peu convenu, pour mettre en évidence quelques « non-dits » méthodologiques.

Summary

Methodology consitutes the study of methods considered as « manners of leading its thought, of establishing or of showing truths according to certain principles and with a certain order ». We will see that medical « truths », knowledge, could be of an extremely different nature: factual knowledge is a simple observation, often with operational vocation; theoretical knowledge belongs to a coherent whole, it fascines and facilitates our representations of pathologies; the causal knowledge is often the subject of extreme interrogations in medical practice; it is of primarily metaphysical nature. These distinctions are not without consequence in methodological terms. Theories are useful because they facilitate the choice of the assumptions to be tested, the variables to be measured and the interpretation of the results of experiments. The risk however exists to be unable to think apart from their framework, and thus to neglect any source of knowledge which they could not integrate. The cause is inaccessible to science; it will be necessary, in practice, to be limited to the search for variables of upstream on which action of prevention or cure could be undertaken. To establish, to show this medical knowledge, it is necessary to resort to the scientific method by formulating refutable assumptions by reproducible experiments, and that this process involves a belief in the found results. It appears thus that for the same experimental plan, the level of belief and thus the level of proof of the results is a function of the type of measurement carried out (commonplace measurement or of high technicality), of the medical discipline and the media character or not of studied pathology, even of stakes of being able. The medical reproducibility of experiments is problematic since results of these experiments are in general random. This results from the considerable number of variables entering in the determination of human pathologies, and that the majority of these variables are either unknown, or impossible to control. Randomisation makes it possible to give a probabilistic character to this risk: it is then possible to resort to mathematics to carry out statistical tests, to calculate confidence intervals and to profit from a possibility of making inferences with a known margin of error. In the absence of randomisation, it is possible to resort to models; they imply however the knowledge of statistical methods, reserve, experience and a great intellectual honesty during the interpretation of the results, finally a replication of the experiments is here, more than elsewhere, essential.