TitleSubgroup analysis as a source of spurious findings: an illustration using new data on alcohol intake and coronary heart disease.
Publication TypeJournal Article
Year of Publication2015
AuthorsBell, S, Kivimäki, M, Batty, GD
JournalAddiction
Volume110
Issue1
Pagination183-4
Date Published01/2015
ISSN1360-0443
KeywordsAlcohol Drinking, Alcohols, Coronary Disease, Humans, Risk Factors
Abstract
 

Ware & Munafò 1 discuss the important topic of chasing statistical significance and note that unplanned post‐hoc analyses do not equate to serious scientific misconduct. We agree, but want to emphasize that the practice can, none the less, be very misleading—particularly when the unplanned analyses become the focal point of a paper 2. With ever‐increasing pressure to publish, the practice of taking well‐established associations and examining them within strata of a third variable, also known as subgroup analyses, is becoming popular 34, yet the social or biological rationale for anticipating that the relationship will differ between subgroups is rarely stated 5. Ultimately, post‐hoc subgroup analyses can lead to spurious conclusions being drawn 67 that may mislead public health and clinical policy 8.

DOI10.1111/add.12708
Alternate JournalAddiction
PubMed ID25515832
PubMed Central IDPMC4273867
Grant ListR01 HL036310 / HL / NHLBI NIH HHS / United States
AG13196 / AG / NIA NIH HHS / United States
K013351 / / Medical Research Council / United Kingdom
R01 AG034454 / AG / NIA NIH HHS / United States
R37 AG013196 / AG / NIA NIH HHS / United States
RG/07/008/23674 / / British Heart Foundation / United Kingdom
HL036310 / HL / NHLBI NIH HHS / United States
309337 / / European Research Council / International
MR/K013351/1 / / Medical Research Council / United Kingdom
R01 AG013196 / AG / NIA NIH HHS / United States
AG034454 / AG / NIA NIH HHS / United States