Rxivist logo

Conditioning on a collider may induce spurious associations: Do the results of Gale et al. (2017) support a protective effect of neuroticism in population sub-groups?

By Tom G Richardson, George Davey Smith, Marcus R. Munafò

Posted 11 Dec 2017
bioRxiv DOI: 10.1101/231969

It was reported earlier this year that neuroticism may have a protective effect on all-cause mortality among certain population sub-groups. In this commentary, we have examined whether "collider bias" may influence findings such as this, using large-scale data from the UK Biobank study. This was undertaken by evaluating the relationship between neuroticism and risk factors for mortality, before and after conditioning on self-reported health. Illustrations of the detrimental effect of collider bias were particularly evident when investigating the association between neuroticism and body mass index, lung function, cancer and diabetes. These results serve as a cautionary note that, while large cohort studies provide unparalleled power to detect novel findings, the ability to identify ever smaller effect sizes increases the risk of relatively weak biases influencing results.

Download data

  • Downloaded 558 times
  • Download rankings, all-time:
    • Site-wide: 58,118
    • In epidemiology: 2,667
  • Year to date:
    • Site-wide: 152,120
  • Since beginning of last month:
    • Site-wide: 113,835

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide