Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
The authors respond:
We have no substantial disagreement with the points raised by Dr. Neutra.1 On the other hand, we think the letter by Mejía-Aranguré2 incorrectly cites our work and exemplifies ideas we criticized in our commentaries.3,4
We never proposed that risk-factor epidemiology not deal with theory, only that it is counterproductive to pressure authors to deal with theory. Such pressure ignores historic lessons that theory is empty without accurate observations, and that division of labor between observers and theoreticians can be fruitful. Consider the 16th-century observations by Tycho Brahe on planetary motions. Their accuracy, not any theory Tycho offered, gave them scientific value, for that accuracy enabled Kepler to show the motions were elliptical. Our case study argued that observations free of theoretical window dressing can be a boon to epidemiology as well. Although theory motivates choice of observations, that theory may disappear (eg, many nutritional studies were motivated by now-obsolete theories about antioxidants and health), yet their data remain valuable to the extent that they are free of error and bias. Theories are disposable, data are not; and accuracy can give observations an enduring value even when the theoretical content (which can only be evaluated relative to current theories) is ephemeral.
Null associations are not equivalent to falsifying observations. An elaborate theory will predict some associations as null and others as nonnull; failure of observations to match predictions is theory refutation. Regardless, not all philosophers agree that evidence against a theory is “more relevant”; some even eschew epistemological rules.5
Worrying about “if a theory is never proposed” is empty. Observations demand explanation and they propel theorization, including theorization about biases.6 There is never a theory shortage, for theorization is cheap and seems a human compulsion (eg, every P value invokes the theory that the association was the result of chance alone). In contrast, accurate epidemiologic data are expensive to gather and usually in short supply.
Finally, divisions between “basic” and public health research are determined by current theory and technology; for example, today's basic genetic research becomes relevant to health as screening tests and therapies develop. Our commentaries3,4 concerned only articles about existing data. Proposal evaluation is different; funders should consider use from a cost-benefit perspec-tive7 without expecting definitive answers from single studies. Confronted with knowledge limitations and overblown utility claims, funders can demand that observations be made as accurately as feasible, so that if the study fails to meet proposal goals, the data may still have value (possibly for unforeseen questions).
Departments of Epidemiology and Statistics, University of California Los Angeles, Los Angeles, CA, firstname.lastname@example.org
Jose Esteban Castelao
Department of Preventive Medicine, University of Southern California, Los Angeles, California
1. Neutra RR.‘Black-box’ epidemiology [Letter]. Epidemiology
2. Mejía-Aranguré JM, Fajardo-Gutiérrez A, Ortega-Alvarez M. ‘Black-box’ epidemiology [Letter]. Epidemiology
3. Greenland S, Gago-Dominguez M, Castelao JE. The value of risk-factor (‘black box’) epidemiology. Epidemiology
4. Greenland S, Gago-Dominguez M, Castelao JE. Authors’ response. Epidemiology
5. Feyerabend P. Against Method
, 3rd ed. London: Verso; 1993.
6. Maclure M, Schneeweis S. Causation of bias: the episcope. Epidemiology
7. Phillips CV. The economics of ‘more research is needed’. Int J Epidemiol