Causation and models of disease in epidemiology

Abstract

Nineteenth-century medical advances were entwined with a conceptual innovation: the idea that many cases of disease which were previously thought to have diverse causes could be explained by the action of a single kind of cause, for example a certain bacterial or parasitic infestation. The focus of modern epidemiology, however, is on chronic non-communicable diseases, which frequently do not seem to be attributable to any single causal factor. This paper is an effort to resolve the resulting tension. The paper criticises the monocausal model of disease, so successful in the nineteenth century. It also argues that a multifactorial model of disease can only be satisfactory if it amounts to more than a mere rejection of the monocausal model. A third alternative, the contrastive model, is proposed and defended on the grounds that it links the notions of disease and of general explanation, while avoiding the philosophical naiveties and practical difficulties of the monocausal model.

Introduction

Two conceptual questions currently face epidemiology, both relating to causation. First, how should it handle certain diseases, which appear to be etiologically more complex than the infections and deficiencies by which epidemiology made its name? In particular, chronic non-communicable diseases (CNCDs) account for a larger proportion of deaths, at least in the industrialised world, than they did in 1900 (Rockett, 1999, p. 8), and attract more epidemiological attention. Yet they often do not seem susceptible to definition in terms of any one causative agent: their etiology is typically complex.

Second, how should epidemiology respond to newly identified causes of disease? Epidemiology has moved beyond obvious environmental causes of illness (such as prolonged extreme cold) and uncovered increasingly complex and sometimes surprising environmental causes of disease. And in place of the old notion of a ‘constitution’, the discipline has had to grapple with a newly discovered category of cause: genetics. The increased depth and complexity of our knowledge of both genetic and environmental determinants of health places pressure on aspects of the conceptual framework of epidemiology: in particular, on the way it thinks about disease causation.

Devising a conceptual framework for thinking about disease causation has proved astonishingly difficult. On the one hand, the early history of epidemiology appears to attest to the power of insisting that every disease has one cause that is necessary and, in limited circumstances, sufficient for the disease. I call this way of thinking about disease etiology the monocausal model of disease. This model suits infectious diseases such as TB and cholera well, along with parasitic infestations and diseases of deficiency. On the other hand, the monocausal model is a terrible fit for CNCDs such as lung cancer or diabetes. It is theoretically possible that a condition like diabetes has a single necessary and, in some circumstances, sufficient cause, which we have not yet discovered. But surely, it is also a theoretical possibility that there is no cause for diabetes satisfying that description. And even if there is, it is not clear how insisting that there must be such a cause helps us achieve public health or clinical goals, if we don’t know what it is. The causes that we are able to identify are causal risk factors: neither necessary nor sufficient. These are all we have to work with. Accordingly, a view of disease as multifactorial now dominates epidemiology. But this is not an entirely happy situation, because it fails to mark what looks like a real etiological difference between diseases like cholera and conditions like lung cancer. The monocausal model has had some striking successes in the history of epidemiology, and these successes are left unexplained by the mere assertion that disease causation is multifactorial. Unless we can explain the successes of the monocausal model in terms of modern multifactorial thinking, there is a risk of throwing the baby out with the bathwater.

In this paper I want to address the tension that arises between monocausal and multifactorial models of disease. Both are, to some extent, rational reconstructions of positions that are implicit in the epidemiological literature. There have been very few (if any) attempts to lay out these two ways of thinking about disease causation, in a fully explicit and philosophically rigorous manner. Accordingly there is a danger of attacking straw men. It should be understood that these models of disease, as I state them, are attempts to make explicit ways of thinking that are implicit: so I refrain from attributing the result of this exercise as an opinion of any historical or contemporary figure. Nevertheless, I do think—and will argue—that these ways of thinking are present in various extant efforts to conceptualise disease causation. Once I have assessed the strengths and weaknesses of these two models of disease, I will propose a ‘contrastive’ model, which attempts to preserve the strengths of the monocausal model within a multifactorial framework.