|
Epi Wit & Wisdom Articles
Epidemiology Department Tackles
Science Article (3 of 6)
Special report by Jonathan Samet
The Department of Epidemiology
of the Johns Hopkins University School of Hygiene and Public Health
discussed the July 1995 Science article by Gary Taubes in one of the
weekly departmental seminars. There was great interest in the article
and a very lively discussion. It was focused around 10 points of
criticism of epidemiology that could be identified in the article.
They are:
1) Epidemiologic studies are
conflicting.
2) Epidemiologic studies
exaggerate risks.
3) Epidemiologic studies cannot
address “weak risks.”
4) Epidemiology is an
observational science.
5) The randomized trial is the
“best” design.
6) Bias is a plague in
epidemiologic studies.
7) Exposure assessment is an
obstacle.
8) Case-control studies are
inherently flawed by the difficulty of selecting controls.
9) Epidemiologic studies can
only address relative risks above 3 - 4.
10) Even consistency does not
necessarily help in interpreting evidence.
The seminar participants noted
that these were not newly identified limitations of epidemiology. Most
of the attendees were surprised by the article’s pervasively critical
tone and some by its publication. Most of the points of concern have
long been recognized and have already been topics for discussion among
epidemiologists.
The observational nature of much
epidemiologic research is evident, but the non-experimental nature of
this research does not necessarily mean that the randomized trial is
the design of choice.
The topics of bias and control
selection have received substantial attention by epidemiologists, as
have methods for exposure assessment. In fact, the new field of
environmental exposure assessment now complements environmental
epidemiology. The article and a number of the quoted epidemiologists
cited arbitrary lower bounds of relative risk, below which the
findings of epidemiologic studies lose credibility. While uncontrolled
confounding or other forms of bias have potentially more serious
effects at lower levels of relative risk, some exposures to risk
factors for disease would be expected to have effects that might be
judged as “weak.” In assessing the credibility of lower levels of
relative risk, consideration should be given to the level of risk
anticipated on a biological basis and not to arbitrary boundaries.
The participants acknowledged
that the article had elements of “truth” about epidemiology and
epidemiologists. Interpretation of epidemiologic data may be
challenging, particularly in investigating poorly measured risk
factors for diseases having multifactorial etiology, e.g. diet and
cancer. There needs to be more effective communication with the media
and the public as the findings of studies are reported. These skills
and an understanding of the policy implications of epidemiologic
research might become part of training in epidemiology.
The seminar ended with a
discussion of next steps. How should the field respond to the article?
Participants offered that little would be gained by responding in
letters to Science. There was consensus that advocacy for the field by
professional organizations was needed.
Published February 1996 v
|
