Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Correspondence
  • Published:

Techniques and Methods

Selection bias can creep into unselected cohorts and produce counterintuitive findings

International Journal of Obesity volume 45, pages 276–277 (2021)Cite this article

Subjects

We read with interest the study by Fekri et al. [1], “Association of body mass index with life expectancy with and without cardiovascular disease” (CVD). This letter demonstrates that some of their findings may be the result of selection bias, even though the study involved a large, seemingly unselected population-based cohort. Selection bias can occur not only during study design or recruitment, but also through restriction or conditioning in the analysis [2, 3].

The study findings can be summarized as follows:

  1. 1.

    Obesity led to earlier CVD—This finding was expected as CVD is a well-known consequence of obesity.

  2. 2.

    Among those with CVD, obesity was associated with a longer life—This finding may seem surprising but is consistent with a non-causal inverse association that develops due to collider stratification (a form of selection bias). Here, obesity becomes inversely associated with other (more dangerous) causes of CVD when analyses condition on CVD [4, 5].

  3. 3.

    Among those without CVD, obesity was associated with a longer life—This finding appears counterintuitive at first. However, as shown below, this outcome can also be explained through collider stratification bias.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Causal diagram with CVD being the result of genes and weight status, and with independence between genes and weight status.

References

  1. Fekri N, Khaloo P, Ramezankhani A, Mansournia MA, Azizi F, Hadaegh F. Association of body mass index with life expectancy with and without cardiovascular disease. Int J Obes (Lond). 2020;44:195–203. https://doi.org/10.1038/s41366-019-0464-3

    Article  Google Scholar 

  2. Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004;15:615–25. https://doi.org/10.1097/01.ede.0000135174.63482.43

    Article  PubMed  Google Scholar 

  3. Infante-Rivard C, Cusson A. Reflection on modern methods: selection bias-a review of recent developments. Int J Epidemiol. 2018;47:1714–22. https://doi.org/10.1093/ije/dyy138

    Article  PubMed  Google Scholar 

  4. Banack HR, Stokes A. The ‘obesity paradox’’ may not be a paradox at all. Int J Obes. 2017;41:1162–3. https://doi.org/10.1038/ijo.2017.99

    Article  CAS  Google Scholar 

  5. Stovitz SD, Banack HR, Kaufman JS. Structural bias in studies of cardiovascular disease: let’s not be fooled by the “obesity paradox”. Can J Cardiol. 2018;34:540–2. https://doi.org/10.1016/j.cjca.2017.10.025

    Article  PubMed  Google Scholar 

  6. Luque-Fernandez MA, Schomaker M, Redondo-Sanchez D, Sanchez Perez MJ, Vaidya A, Schnitzer ME. Education Corner Educational Note: paradoxical collider effect in the analysis of non-communicable disease epidemiological data: a reproducible illustration and web application. Int J Epidemiol. 2019;48:640–53. https://doi.org/10.1093/ije/dyy275

  7. O’Brien EC, Thomas LE. Untangling the paradox: obesity as prognostic marker in prevalent cardiovascular disease. Am Heart J. 2016;172:170–2. https://doi.org/10.1016/j.ahj.2015.11.005

    Article  PubMed  Google Scholar 

  8. Stovitz SD, Banack HR, Kaufman JS. Paediatric obesity appears to lower the risk of diabetes if selection bias is ignored. J Epidemiol Community Health. 2018;72:302–8. https://doi.org/10.1136/jech-2017-209985

    Article  PubMed  Google Scholar 

  9. Stensrud MJ, Valberg M, Aalen OO. Can collider bias explain paradoxical associations? Epidemiology. 2017;28:e39–40. https://doi.org/10.1097/EDE.0000000000000653

    Article  PubMed  Google Scholar 

  10. Day FR, Loh PR, Scott RA, Ong KK, Perry JRB. A robust example of collider bias in a genetic association study. Am J Hum Genet. 2016;98:392–3. https://doi.org/10.1016/j.ajhg.2015.12.019

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Department of Family Medicine and Community Health, University of Minnesota, Minneapolis, Minnesota, USA

    Steven D. Stovitz

  2. Department of Epidemiology and Environmental Health, School of Public Health and Health Professions, University at Buffalo, The State University of New York, New York, New York, USA

    Hailey R. Banack

  3. Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Quebec, Canada

    Jay S. Kaufman

Authors
  1. Steven D. Stovitz
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Hailey R. Banack
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Jay S. Kaufman
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Steven D. Stovitz.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Stovitz, S.D., Banack, H.R. & Kaufman, J.S. Selection bias can creep into unselected cohorts and produce counterintuitive findings. Int J Obes 45, 276–277 (2021). https://doi.org/10.1038/s41366-020-00720-2

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41366-020-00720-2

This article is cited by

Search

Advanced search

Quick links