Wing: Descriptive Epidemiology by Any Other Name...

[dj_strom@ccmail.pnl.gov (Daniel J Strom)]

In 1877 Henle and Koch first addresssed association and causation (Evans 
1976), and their work was revisited two decades ago (Rothman 1976).  Sir 
Austin Bradford Hill published an influential work on association and 
causation (Hill 1965).  Another perspective can be found in the work of 
Susser (1991).  Canadian and U.S. leaders addressed the quality of 
epidemiologic evidence under the title "Hierarchy of Evidence" (U.S. 
Preventive Services Task Force 1989).  Most recently, the Federal Focus 
expert panel explained why descriptive epidemiology studies don't get 
much respect among risk analysts when they try to come to quantitative 
conclusions (Graham et al. 1996).  

I have always been critical of the application of descriptive 
epidemiology (e.g., ecologic studies) to quantitative problems (Strom 
1997, 1991a).  "Descriptive studies are generally viewed as useful for 
identifying or formulating causal hypotheses, but not a sufficient to 
test such hypotheses, because they lack data on individuals, such as 
individual exposures, potential confounding exposures, factors affecting 
individual susceptibility, and potential biases.  In contrast, studies 
generally termed 'analytic' aim to establish risk factors for 
populations and individuals by ascertaining individual exposures and 
controlling for other variables such as gender, age, race, or exposure 
to other agents that could affect risk estimates independently 
(potential 'confounders'), potential study biases, and variations in 
host susceptibility.  There are two main types of analytic epidemiology: 
case-control and cohort studies..."  This quote is from a new book 
written by an expert panel of risk assessors in 1995 entitled 
"Principles for Evaluating Epidemiologic Data in Regulatory Risk 
Assessment"  (Graham et al. 1996).  The panel was comprised of an 
international group (mostly from the USA, however) of well-respected, 
middle-of-the-road risk assessors from universities, governments, and 
industry groups.  

Steve Wing and colleagues have published a reanalysis of TMI health 
effects data (Wing et al. 1997).  Both the new Wing study and the work 
of B.L. Cohen (Cohen 1995) are in a category of descriptive 
epidemiology, as opposed to analytical epidemiology.  

Whether descriptive or analytic, virtually all occupational and 
environmental epidemiology studies are "observational" as opposed to 
"experimental" (a.k.a. clinical or interventional) studies.  Since human 
experimentation, outside of closely supervised clinical trials, is out 
of the question, we are left with observational study designs which, 
unfortunately, are not the most cogent designs because of uncontrolled 
factors.  Neither the Wing TMI study nor Cohen's study are 
"experiments," but rather compilations and analyses of whatever data are 
available.

If you are upset by Wing yet celebrate Cohen, I ask that you examine why 
descriptive studies are compelling in one case and not in the other.  To 
me, the bottom line is that neither have data for individuals, neither 
has meaningful control for confounders and biases, and no amount of 
statistical analysis will change that.  Both fail to meet many of the 
criteria presented by leading risk analysts.

I am reminded of the brouhaha about Steve Wing and co-workers' earlier 
study of ORNL workers (Wing et al. 1991), which at least was an analytic 
cohort study.  Lest I be branded as being on one side of the the issue 
of radiation risks, I note that my reply to Wing (Strom 1991b) included 
direct criticism of the methods as well as a reminder to consider the 
Bradford Hill criteria (Hill 1965) for interpreting an association as 
causal as reiterated by the Expert Panel (Graham et al. 1996).

As an illustration of the Bradford Hill criteria, I offer a 
tongue-in-cheek quote from my colleague Dwight Underhill: "In the 
winter, I wear galoshes.  In the winter, I get colds.  Therefore, 
galoshes cause colds."  Association?  22 standard deviations, I'd guess. 
Causation?  Not by criteria I use.

I must also confess that I do not form my opinions on the basis of books 
published by what National Public Radio calls "the Libertarian Cato 
Institute," which published Steve Milloy's "Junk Science" book.  At the 
associated web site, one finds ecologic studies celebrated if they 
support deregulation, no effects, or hormesis; and denounced if they 
support regulation, or harmful effects of some agent or other.  Again, 
these are not criteria I use to judge the weight of epidemiologic 
evidence.

References

Cohen, B.L.  Test of the Linear-No Threshold Theory of Radiation 
Carcinogenesis for Inhaled Radon Decay Products.  Health Physics 
68(2):157-174; 1995.

Evans, A.S.  Causation and Disease: The Henle-Koch Postulates Revisited. 
 Yale Journal of Biology and Medicine 49:175-195; 1976.

Graham, J.D.; Koo, L.C.; Paustenbach, D.J.; Wynder, E.L.; Ashby, J.; 
Carlo, G.; Cohen, S.M.; Evans, J.S.; Holland, W.; Matanoski, G.M.; 
North, G.W.; Pershagen, G.; Schlesselman, J.J.; Starr, T.B.; Swenberg, 
J.A.; Teta, M.J.; Wichmann, E.; Williams, G.M.; Kelly Jr., W.J.; 
Auchter, T.G.; Landeck, S.; Ploger, W.D.  Principles for Evaluating 
Epidemiologic Data in Regulatory Risk Assessment. Washington, DC: 
Federal Focus, Inc. 1996.

Hill, A.B.  The Environment and Disease: Association or Causation?  
Proceedings of the Royal Society of Medicine 58:295-300; 1965.

Rothman, K.J.  Causes.  American Journal of Epidemiology 104(6):587-592; 
1976.

Strom, D.J.  The Ecologic Fallacy.  Health Physics Society Newsletter 
19(3):13; 1991a.

Strom, D.J.  A Critique of "Mortality Among Workers at Oak Ridge 
National Laboratory".  Nuclear News 34:67-74; 1991b.

Strom, D.J.  Radon Study Shows Little Correlation.  Letter.  Health 
Physics 72(3):488-489; 1997.

Susser, M.W.  What is a cause and how do we know one?  A grammar for 
pragmatic epidemiology.  American Journal of Epidemiology 133:635-648; 
1991.

U.S.Preventive Services Task Force.  Guide to Clinical Preventive 
Services. Baltimore: Williams & Wilkins; 1989.

Wing, S.; Shy, C.M.; Wood, J.L.; Wolf, S.; Cragle, D.L.; Frome, E.L.  
Mortality Among Workers at Oak Ridge National Laboratory: Evidence of 
Radiation Effects in Follow-Up Through 1984.  Journal of the American 
Medical Association 265(11):1397-1402; 1991.

Wing, S.; Richardson, D.; Armstrong, D.; Crawford-Brown, D.J.  A 
Reevaluation of Cancer Incidence Near the Three Mile Island Nuclear 
Plant: The Collision of Evidence and Assumptions.  Environmental Health 
Perspectives 105(1):52-57; 1997.


The opinions expressed above are my own, and have not been reviewed or 
approved by Battelle, the Pacific Northwest National Laboratory, or the 
U.S. Department of Energy.

Daniel J. Strom, Ph.D., CHP
Staff Scientist
Health Protection Department K3-56
Pacific Northwest National Laboratory
Battelle Boulevard, P.O. Box 999
Richland, WA 99352-0999 USA
(509) 375-2626
(509) 375-2019 fax
dj_strom@pnl.gov