about the history of epidemiologic methods
I'm trying to find out who first used the term "Relative Risk" in epidemiology. Can you help?
Bruce V. Stadel, MD, MPH (FDA)
To my knowledge, the first usage of the term RR as a ratio of risks
is in Cornfield J et al's classic paper: Smoking and lung cancer.
JNCI 1959;22:173?203. MacMahon, Pugh and Ibsen's Epidemiologic method
in 1960 is the first textbook using the term RR and defining it.
I cannot guarantee that someone did not use the term before Cornfield
and his colleagues, but I would be surprised if the term was commonly
used before 1959. In 1951, Cornfield was using the term relative
frequencies. I have not found the term in the first edition of Morris's
textbook Uses of epidemiology from 1957. But it suddenly appears
in many publications in 1959. Interestingly, many (if not most publications)
use it to define odds ratio (e.g., Dorn HF. Some problems arising
in prospective and retrospective studies of the etiology of disease.
NEJM 1959; 261:571?9, or in the classic 1959 Mantel Haenszel paper).
The hypothesis that Cornfield et al's paper was the first to use
the term RR is very plausible for several reasons: 1) the list of
co?authors is impressive: Haenszel, Hammond, Lilienfeld, Shimkin
and Wynder. They were the leaders of US epidemiology in the fifties.
2) the paper was probably the first theoretical explanation of the
rationale for using both absolute and relative measures of effect
(more or less the explanation you gave yourself in the intro of
OC and CVD 1981, part 1). One aspect that I find particularly interesting
is that the term appears in the explanation of a phenomenon that
has been discussed by others later under the label of the “low
risk group” approach, but to my knowledge never referred to
this landmark publication, that is, that the apparent RR can be
stronger when measured in a low risk than in a high risk population:
"If two correlated agents, A and B, each increase the risk
of disease, and if the risk of the disease in the absence of either
agents is small, then the apparent relative risk (!) for A, r, is
less than the risk for A in the absence of B". (AM)
* I will be happy to post other
contributions on this topic