TOPIC: Multicausality: Confounding Assignment
Confounding is a distortion of the association between an exposure and an outcome by an extraneous, third variable known as the confounder. Unlike information and selection biases which can be introduced by the subjects or by the investigator, confounding is an actual relationship that ought to be adjusted both in the design of study and analysis. When present, confounding obscures the actual effect of an exposure on outcome. In public health, it results in a biased estimate of the effect of exposure on disease.
Investigators should therefore focus on recognizing, preventing, and controlling confounding for accurate and reliable estimates. The effect of a cofounder can be reduced by adjusting the study design. But to correctly do so, an investigator must first be able to identify whether a given variable is a possible cofounder. For a variable to be classified as a potential cofounder, it must satisfy all these three conditions.
I. It must have an association with the disease, that is, it must be related to the risk factor of interest as well as the outcome.
II. It must be associated with exposure, or rather be unequally distributed between the different exposure groups being considered.
III. It cannot be an effect of the exposure. This means that it should not be an intermediary step in the causal pathway from the exposure of interest to the outcome of interest.
Let’s take the case of alcohol consumption. It is widely acknowledged that one main benefit of modest alcohol consumption is the reduced risk of coronary heart disease. Alcohol works by increasing the levels of HDL, or the good cholesterol which is associated with a low risk of coronary heart disease. In this case, HDL levels are not a cofounder of the association between alcohol and heart disease since it is part of the mechanism by which alcohol produces a beneficial effect. Because higher HDL levels result from taking alcohol and also part of the mechanism by which lower risk of heart disease is achieved, then it is not a cofounder.
However, because most diseases have multiple risk factors, then it is also not unusual to have several cofounders. A cofounder can also be another risk factor for the disease. For instance, in a hypothetical cohort study seeking to establish the relationship between exercise and heart disease, age is a cofounder because it is also a possible risk factor for heart disease.
Preventative factors of a disease can also qualify as cofounders. For instance, if people who exercise frequently also took aspirin, and aspirin lowers the risk of heart disease, then aspirin will be a cofounder since it tends to magnify the benefits of exercising.
Lastly, a cofounder can also be an indicator for another cause of the disease. A factor like the socioeconomic status of an individual can qualify as a cofounder because it is a pointer for a series of factors associated with an elevated risk of heart disease.
Fuller, J. (2019). The confounding question of confounding causes in randomized trials. The British journal for the philosophy of science, 70(3), 901-926.
Greenland, S. (1989). Modeling and variable selection in epidemiologic analysis. American journal of public health, 79(3), 340-349.
Santos, S., Zugna, D., Pizzi, C., & Richiardi, L. (2019). Sources of confounding in life course epidemiology. Journal of developmental origins of health and disease, 10(3), 299-305.

Published by

Essays

View all posts