I wish to turn attention to an area of epidemiological research that is increasingly important, the concept of causality. In epidemiology, an intervention or public health action is based on the presumption that associations seen in various studies are causal, not due to bias or some other spurious research. When we say causal, we mean that the cause of, say, a disease is “an event, condition, characteristic, or combination of factors that plays an important role in the development of disease or health condition." (1) Of course, we realize that demonstrating causality can be exceedingly difficult; how do we definitively know that an association is causal?
Initial descriptions of causality were developed by Henle and Koch (2), relating more to identifying infectious agents in disease outbreaks. Here, the agent had to present in each case of the disease when cultured, could not be found in other diseases, could reproduce the disease in question when isolated and injected into animals, and must then be recovered from the infected animal. But this approach does not work as well for modern public health, where exposures to environmental factors may take years to occur. Thus, Bradford Hill (3) came up with his criteria for causality, which have been modernized into these 6 key factors:
1. Consistency: the association is seen in different settings and populations. Thus, the likelihood of a causal relation increases as the number of studies with similar findings occurs.
2. Strength: This is defined by the size of the relative risk estimate; thus, the likelihood of a causal relation increases as the summary relative risk estimate increases (bringing in a concept we have previously explored in the context of evidence-based care). The larger the effect estimate, the less likely it is due to bias or chance.
3. Temporality: that is, there has to be demonstrated an exposure to the risk prior to the development of the condition of interest. This is a critically important aspect of causality.
4. Dose-response relationship: This is defined as the observed relationship between the dose of the exposure and the magnitude of the relative risk estimate. Put another way, what this means, is that the larger the exposure (intensity or time), the greater the chance (risk) of developing the condition of interest.
5. Biological plausibility: What we know about the mechanism of action for a given risk factor and the disease outcome. What do we know about why smoking leads to cancer, for example? If we can develop a biological rationale for why the exposure leads to the condition of interest, it strengthens the likelihood that a causal relation is present.
6. Experimental evidence: This is a bit more technically sophisticated; here, defined as the presence of findings from a prevention trial in which we remove the exposure from randomly assigned individuals. If we see a reduction in the condition of interest, it strengthens the likelihood that a causal relation exists.
Our challenge becomes removing all potential confounders from consideration, and given this one can se why developing causal relations can be so difficult. An example was a study showing that coffee drinkers had a higher risk of cancer. Of course, it was not the coffee drinking which was involved; it turned out that there were more smokers among the coffee drinkers and that was what led to higher cancer rates. Another challenge is to apply this to chiropractic theory and methods, including subluxation research.
1. Brownson RC, Baker EA, Leet TL, Gillespie KN. Evidence-based public health. New York; Oxford University press, 2003:30
2. Last JM, ed. A dictionary of epidemiology, 4th edition. New York; Oxford University Press, 2001
3. Hill AB. The environment and disease: association or causation? Proc Royal Soc Med 1965;58:295-300