Epidemiology for the practicing veterinarian (Proceedings)

In the era of evidence-based medicine – or the “…conscientious, explicit, and judicious use of current best evidence in making  decisions about the care of individual patients” [1] – it is critical that practitioners have a strong epidemiological foundation upon which clinical experience and best available external evidence can be integrated.  Implied in this is that the best evidence is clinically relevant to the case at hand.  By using epidemiology, practitioners can evaluate the quality of the available evidence and apply it appropriately to manage individual patients and patient populations. 

Disease does not occur randomly in populations – So why are some more likely to develop disease than others?  In the occurrence of health and disease one must consider the interaction among the agent, the host, and the environment (i.e., the epidemiological triad) and how differential exposure to disease “causes” may account for variability in disease occurrence.  By looking at patterns of disease in populations, making comparisons between groups, and describing changes over time, we can begin to understand factors contributing to the “cause” of disease.  This is critically important as we can only prevent disease by removing exposure to causal factors.

Causal relationships are not always simple – many can be quite complex with multiple contributing factors (i.e., causal factors).  To gain an understanding of such complex causal relationships we often turn to models such as the necessary-sufficient-component causes model, path models, Venn diagrams, and statistical models.  Each contributes to our understanding in different ways.  The necessary-sufficient-component causes model suggests that there are factors or component causes, that together provide the complement of factors contributing to the occurrence of disease [2].  Among those are causes that are necessary for the occurrence of disease (these are always present when disease occurs) and a constellation of factors which interact to produce a sufficient cause (a group of factors that when present results in disease).  Alternatively, while path models provide some insight into the sequence of occurrence for the causal factors contributing to disease development, Venn diagrams show the magnitude and interaction of these factors in diagrammatic form.  Finally, statistical models can help us understand which factors may be associated with the occurrence of disease and provide some insight into its magnitude of effect.

While we spend a significant amount of time on the “cause” of disease, causation can never definitively be proven.  Rather we can establish the presence of causal relationships through the use of Koch's Postulates, Hill's Criteria, statistics, and scientific judgment.  Koch's Postulates are based on the premise that one agent leads to one disease and are stipulated as follows:  1) the agent is found in all cases of disease; 2) the agent is not found in other individuals as a non-pathogen; 3) it must be possible to produce a pure, sustainable culture of the agent; and 4) it must be possible to produce the same disease in a susceptible host [3].  While these postulates may prove useful for diseases such as rabies, they are too limited for more complex causal relationships which may have many different sufficient causes such as bovine respiratory disease.  Hill's Criteria were put forward by Sir Austin Bradford Hill as a means for identifying causal relationships [4].  The criteria include the following:  1) time sequence (i.e., the exposure must precede disease); 2) strength of association (exposure is more common in disease and there is a large difference in disease risk); 3) consistency of findings (relationships demonstrated in different populations using different study designs are more likely to be causal); 4) biological gradient or dose effect (changes in exposure lead to changes in disease risk); 5) specificity of association (one exposure – one disease); 6) coherence with established facts; and 7) biological plausibility.  We reach for statistics when defining the role of chance, i.e., to be objective in our confidence in the data;  trying to answer the question of was the outcome of the study a coincidence or did a causal mechanism exist?  Statistics uses both the magnitude of the difference and precision in the estimates to draw conclusions about the existence of a causal relationship. 

Applying best-practices to the practice of veterinary medicine not only requires practitioners to be aware of current evidence, but requires them to be able to assess the quality of that evidence and determine if it is applicable to the patient standing before them.  This demands a basic understanding of study design, measures of disease occurrence (e.g., incidence and prevalence), and measures of association (e.g., relative risks and odds ratios).  Through using the tools in their epidemiological toolbox and critically evaluating the literature, practitioners can provide high-quality care to patients and populations alike.

References

Sackett, D.L., Rosenberg, W.M., Gray, J.A., Haynes, R.B. and Richardson, W.S. (1996) Evidence based medicine: what it is and what it isn't. Bmj 312, 71-72.

Rothman, K.J. (1976) Causes. Am J Epidemiol 104, 587-592.

Evans, A.S. (1976) Causation and disease: the Henle-Koch postulates revisited. The Yale journal of biology and medicine 49, 175.

Hill, A.B. (1965) The environment and disease: association or causation? Proceedings of the Royal Society of Medicine 58, 295.