To the Editor:
I find the article on “black-box” epidemiology by Greenland et al1 to be valuable and thought-provoking. What have we achieved by documenting that the many epidemiologic risk factors for renal carcinoma are also associated with lipid peroxidation? Greenland and colleagues suggest 2 possibilities.
First, we can do a Bayes-influenced evaluation of the pattern of data. The experimental pathology has given us a kind of qualitative prior odds that lipid peroxidation causes renal carcinoma. We can combine this information with the qualitative relative likelihood conveyed by the fact that all these risk factors for renal carcinoma also are associated with lipid peroxidation. It would be unlikely to have all those associations under an alternative hypothesis. Thus, our qualitative relative likelihood would be a lot bigger than 1.00, so when multiplied by our prior, our qualitative posterior odds for lipid peroxidation causing renal carcinoma should increase a lot.
Alternatively, if we do not think like Bayesians, we can turn to the “good story” and “Occam's razor” heuristics. These would also lead us to boost our degree of certainty about the lipid peroxidation hypothesis. Furthermore for sociologic reasons this evidence would convince both epidemiologists and experimental pathologists (because scientists put more weight on evidence generated by their own disciplines).
There are 2 other benefits to be derived from the epidemiologic associations. They tell us that the arcane biochemical mechanism may well be at work in the real world. Also, epidemiology tells us about the magnitude of the morbidity that results from the separate causal pathways and the final common lipid peroxidation pathway, providing estimates of the fraction preventable by various kinds of interventions.
In short, the epidemiology puts the biochemistry in its proper community perspective.
Raymond Richard Neutra
Division of Environmental and Occupational Disease Control California, Department of Health Services Oakland, California firstname.lastname@example.org
1. Greenland S, Gago-Domingquez M, Castelao J. The value of risk-factor (‘black-box’) epidemiology. Epidemiology