Decoding health and disease
The purpose of this blog is to present, discuss and share knowledge on topics related to population health and medical research. Blog posts are written mainly by researchers from the Faculty of Medicine.
“Dementia has one surprising risk factor” -Iltalehti 25.2.2023
“An underlying risk factor of stroke may be missed” -Ilta-Sanomat 9.12.2023
These are a couple of tabloid headlines from the past year which include the term “risk factor”. This term also belongs to the vocabulary of medical practitioners and researchers: as of 19th December 2023, the search for the Finnish term “riskitekijä” provides 159 hits in the University of Oulu News database, mostly relating to theses or publications in medical research or epidemiology.
If broken into pieces, it is obviously some sort of a “factor” which is somehow related to the “risk” of an outcome or an event of interest. If we are interested in, say, the occurrence of a disease, then a risk factor must be something to avoid if possible. In this case, a risk factor can be a condition such as smoking, obesity, or a measurement, such as blood cholesterol levels. These three factors are widely known to be associated with multiple diseases and disorders.
However, from the perspective of medicine and public health, it is vital to know in more detail the nature of the relation between the “factor” and the “risk” of the event of interest – whether there is a causal association from the risk factor (cause) to the outcome (consequence), or whether it is merely a correlation (a statistical association) between the risk factor and the outcome. In the case of a causal relationship, it is expected that if the risk factor is successfully avoided, the outcome will be less likely to occur. The same reduction in the outcome risk cannot be directly expected if there is only a non-causal correlation between the risk factor and the outcome. The knowledge on causal relationships is therefore crucial, for example when attempting to prevent a disease.
As a concrete example: Blood cholesterol levels can be crudely divided into “good” HDL-cholesterol and “bad” LDL-cholesterol. According to the current knowledge, high LDL-cholesterol levels have a causal relationship with higher risk of coronary heart disease. Therefore, pharmaceutical drugs that lower LDL-cholesterol often work in reducing the risk of heart disease, and indeed are widely used.
However, while low HDL-cholesterol levels do correlate with the risk of heart disease, there is no conclusive evidence of a causal relationship. Likewise, there is currently no support for drugs that increase HDL-cholesterol levels for heart disease prevention. Despite this, HDL-cholesterol levels can be used in different heart disease risk calculators (https://www.kaypahoito.fi/pgr00069).
The question is: Are both high LDL-cholesterol levels and low HDL-cholesterol levels risk factors for heart disease?
Unfortunately, the answer to this question is somewhat mundane: the meaning of the term “risk factor” depends on how it is defined by its user, and the term is not well-defined even among researchers!
The problem in defining the meaning of a “risk factor” depends on what is of interest: if a researcher wants to predict an individual’s probability of a heart disease, low HDL-cholesterol levels can be considered as a predictive risk factor. However, if the aim of the researcher is to find the reasons for increased risk of heart disease, then low HDL-cholesterol levels are not a risk factor. Of note, low HDL-cholesterol levels do correlate with the disease risk as well as with a causal risk factor of high LDL-cholesterol levels. Therefore, the statistical association between low HDL-cholesterol levels and heart disease risk is partially due to the confounding effect of high LDL-cholesterol levels.
Due to the varying nature in defining risk factors, one might encounter some hilarious claims about risk factors. One example is the rhetorical question posed by Anders Huitfeld in the title of a 2016 paper: “Is caviar a risk factor for being a millionaire?”. If one sees a person ordering a pricey portion of caviar in a restaurant, one can expect this person to be wealthy. Similarly as in the heart disease example, one could create a “risk calculator” for being a millionaire, where the knowledge of a person’s restaurant habits could be used in predicting whether this person is likely to be wealthy or not. Caviar eating can therefore be interpreted as a predictive risk factor for affluence.
However, caviar eating in itself does not make anyone rich, which means that it is not a causal risk factor in this case. It is likely that the direction of causality goes the other way: being rich is likely to make one more liable to order caviar in a restaurant.
Widely used terms can therefore be very ambiguous even among the experts. The responsibility of a more accurate language is, at least partially, down to us researchers (all the main authors of the blog admit of using this vague and sometimes confusing term). It would be good if we researchers were more aware of the difference between “predictive risk factors” and “causal risk factors”.
The authors of Decoding Health and Disease blog would like to wish everyone a relaxing Christmas time and a reasonable amount of caviar for everyone’s Christmas dinner!
Huitfeldt A. Is caviar a risk factor for being a millionaire? BMJ (2016). https://doi.org/10.1136/bmj.i6536.
Shaya GE et al. Coronary heart disease risk: Low-density lipoprotein and beyond. Trends Cardiovasc Med. (2022). https://doi.org/10.1016/j.tcm.2021.04.002.
Main author: Ville Karhunen