Article Text

other Versions

PDF
The possibility of unmeasured confounding variables in observational studies: a forgotten fact?
  1. James Brian Byrd1,
  2. P Michael Ho2
  1. 1Division of Cardiology, University of Colorado, Denver, Colorado, USA
  2. 2Cardiology Section, Denver VA Medical Center, Denver, Colorado
  1. Correspondence to P Michael Ho, 1055 Clermont St. Denver, CO 80220, USA; michael.ho{at}coloradooutcomes.org

Statistics from Altmetric.com

“For a forgotten fact is news when it comes again”–Samuel Clemens

Despite oft-repeated caveats about the proper interpretation of data from observational studies, the implications of such studies are subject to being misunderstood. In recent years, a rich body of observational data has illuminated an association between poor adherence to medication and adverse outcomes. A simple explanation for this relationship is that patients who do not take their medication fail to obtain the beneficial effects of the drugs. However, the confounding of such associations by unmeasured variables is another possibility. In their paper published in Heart, LaFleur et al report on the association between adherence to anti-hypertensive treatment and cardiovascular outcomes and argue that because the benefit found in their observational study did not disappear after controlling for longitudinal blood pressure measurements, unmeasured confounders must be the explanation for the difference found (in press).1 This work reminds us of the need for careful interpretation of findings from observational studies, and, indeed, from studies of all designs. At the time of publication of these data, it is appropriate to reflect upon the implications of this publication, as well as the authors' assumptions and conclusions.

The main strength of the approach taken by Lafleur et al is their rigorous approach in assessing the association between adherence to anti-hypertensive medication and adverse outcomes. The authors used previously validated approaches to measure adherence and examined its association with cardiovascular outcomes after controlling for follow-up blood pressure measurements. In …

View Full Text

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles