Article Text

other Versions

PDF
Placing epidemiological results in the context of multiplicity and typical correlations of exposures
  1. Chirag J Patel1,
  2. John P A Ioannidis2,3,4,5
  1. 1Center for Biomedical Informatics, Harvard Medical School, Boston, Massachusetts, USA
  2. 2Department of Medicine, Stanford Prevention Research Center, Stanford, California, USA
  3. 3Department of Health Research and Policy, Stanford University School of Medicine, Stanford, California, USA
  4. 4Department of Statistics, Stanford University School of Humanities and Sciences, Stanford, California, USA
  5. 5Meta-Research Innovation Center at Stanford (METRICS), Stanford, California, USA
  1. Correspondence to Professor John P A Ioannidis, 1265 Welch Rd, MSOB X306, Stanford, CA 94305, USA; jioannid{at}stanford.edu

Abstract

Epidemiological studies evaluate multiple exposures, but the extent of multiplicity often remains non-transparent when results are reported. There is extensive debate in the literature on whether multiplicity should be adjusted for in the design, analysis, and reporting of most epidemiological studies, and, if so, how this should be done. The challenges become more acute in an era where the number of exposures that can be studied (the exposome) can be very large. Here, we argue that it can be very insightful to visualize and describe the extent of multiplicity by reporting the number of effective exposures for each category of exposures being assessed, and to describe the distribution of correlation between exposures and/or between exposures and outcomes in epidemiological datasets. The results of new proposed associations can be placed in the context of this background information. An association can be assigned to a percentile of magnitude of effect based on the distribution of effects seen in the field. We offer an example of how such information can be routinely presented in an epidemiological study/dataset using data on 530 exposure and demographic variables classified in 32 categories in the National Health and Nutrition Examination Survey (NHANES). Effects that survive multiplicity considerations and that are large may be prioritized for further scrutiny.

  • BIOSTATISTICS
  • Environmental epidemiology
  • Epidemiological methods
  • GENETIC EPIDEM

Statistics from Altmetric.com

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.