Frank Davidoff, MD; Irwin H. Rosenberg, MD
Disclosures: Disclosures can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=L14-0323.
Davidoff F, Rosenberg IH. Association of Dietary, Circulating, and Supplement Fatty Acids With Coronary Risk. Ann Intern Med. 2014;161:454. doi: 10.7326/L14-5018-3
Download citation file:
Published: Ann Intern Med. 2014;161(6):454.
TO THE EDITOR:
The review by Chowdhury and colleagues (1) provides a sobering reappraisal of the widely presumed association between dietary fat and coronary disease. Unfortunately, their otherwise careful analysis uncritically accepts the assumption that the size of an intervention's effect in individual members (or subgroups) of a study population is the same as that of the entire study population; that is, the review fails to avoid the ecological fallacy.
Kent and coauthors (2) identify 2 potentially serious clinical consequences of ignoring the ecological fallacy. Both are due to the inherent risk-based heterogeneity of absolute treatment effects (3), which has been shown to vary as much as 20-fold between study population subgroups with the highest versus lowest baseline risk for adverse outcomes (2). The first problem is failure to recognize that some interventions whose efficacy is statistically confirmed in an entire study population provide no meaningful benefit to sizeable subgroups of that population. For example, warfarin prevents stroke more effectively than aspirin in the overall population of patients with nonvalvular atrial fibrillation, but the subgroup of patients without additional risk factors for stroke does not benefit incrementally from warfarin therapy (2). The second, and opposite, problem is failure to recognize that some interventions provide true benefit in subgroups of a study population even though the intervention is not shown statistically to “work” in the population as a whole. The inclusion of study populations at widely varying baseline risk for adverse coronary events in the review by Chowdhury and colleagues (1) greatly increases the likelihood that its broadly negative conclusion is, at least in part, falsely negative.
Learn more about subscription options.
Register Now for a free account.
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only