Re: Is caviar a risk factor for being a millionaire?
I thank Prof. Porta for his comments. He correctly points out some inaccuracies in the citation for the International Epidemiological Association's "A Dictionary of Epidemiology", I take responsibility for these inaccuracies and apologize to Prof. Porta.
I want to note that I consider the definition in the Dictionary to be sound. If we were only able to get all publishing epidemiologists to adopt this definition, then most of the problems in this article would immediately be solved.
However, I believe it is an empirical fact that many authors use the term "risk factor" in a different sense than that prescribed by the dictionary. For example, under this definition, "family history" is not a risk factor for heart disease. Moreover, if everyone accepted the definition from the dictionary, then the disclaimer "correlation does not imply causation" would be seen as an admission of complete inability to answer research questions about risk factors, and any attempt to study risk factors using prediction models would be immediately rejected, without review, for futility.
As a methodologist trained primarily in causal inference, I recognize that much important research is done to reduce diagnostic and prognostic uncertainty. I worry that any prescriptive attempt to define "risk factor" in a way that excludes this work is unlikely to be acceptable to a large group of practicing scientists, and that such attempts are therefore unable to change how the term is used in practice. This is part of the reason I suggest abandoning the term altogether, and splitting it up into its component pieces.
Competing interests: No competing interests