Morabia Responds to “The Context and Challenge of von Pettenkofer's Contributions to Epidemiology”

In their invited commentary, Oppenheimer and Susser (1) have exquisitely synthesized the issues at stake. They abundantly cite Evans’s book (2), which should belong in every epidemiology curriculum (3). In their last paragraph, Oppenheimer and Susser mention a possible disagreement with me for not having specified that, in von Pettenkofer’s view, interactions could have taken place at both the individual and the societal levels. It was definitely not my intention to, as McMichael (4) would put it, ‘‘imprison’’ von Pettenkofer’s ideas in the proximate, individual level. My paper (5), however, may express more skepticism than that by Oppenheimer and Susser (1) about the extent to which epidemiologic methods can deal with interactions. I believe that attempting to account for too many potential interactions is incompatible with epidemiology. To understand why, let us zoom out from the Koch-von Pettenkofer debate and consider it within a wider historical context of causal thinking about epidemic diseases. The word ‘‘interaction’’ is big in science but trivial in everyday language. In epidemiology, statistics, or system theory, it describes situations of effect amplification (e.g., synergy), contraction (e.g., antagonism) or, more generally, departure from simple combinations of elementary properties of causes. In common language, to interact simply means ‘‘to act together’’ and includes interacting and noninteracting (in the scientific sense of the word) instances. The word is trivial because human intelligence perceives interactions as ubiquitous. On the topic of health, for example, early (traditional) medical theories are replete with interactions. Consider the medical systems of antiquity in ancient Egypt, India, China, and Greco-Roman Europe. With some variations, they were all based on theories linking human health to almost everything else in the universe. If we put aside the poetic feel of these holistic theories, we find that they were refined attempts to explain health and disease as the result of complex interactions. These approaches distinguished microand macrocosmic levels of organization and connected elements, planets, compass points, organs, climates, seasons, colors, tastes, body fluids, facial expressions, and so on by use of rich arrays of correspondence. Interactions, however, are double-edged swords: They lump causes but split effects. There is a density of interactions at which each individual event appears to result from a unique set of causes. Let us refer to this as individualistic thinking. Considering each new case of plague as being determined by a specific set of causes is an obstacle when trying to understand the epidemic’s origin. An individual event, as opposed to the probability of an event in a population, cannot be predicted with certainty and does not provide for comparative causal research. In 1637, the French philosopher René Descartes (1596– 1650) proposed a radically different treatment of interactions in science. The Cartesian method of knowledge acquisition comprised three rules: trust only ‘‘evidence,’’ break up complex systems into simple parts, and ignore interactions (6, pp. 39–40). The approach was in essence reductionist, yet this reduction of complexity made epidemiology possible: A single determinant, isolated from a complex array of causes, could be responsible for many different individual events. Instead of multiple interacting causes that explained an individual event, there could now be one cause that explained a set of events for a multitude of individuals. The Cartesian step liberated probabilistic and population thinking. Group comparison and population thinking, that is, epidemiology (7), applied to simple causal models (e.g., contagious ‘‘cholera poison’’ being transmitted by water) generated more knowledge about epidemic diseases than centuries of holistic thinking had previously uncovered.