[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Views 337
Citations 0
Comment & Response
March NaN, 2016

Nonlinear Exposure-Outcome Associations and Public Health Policy—Reply

Author Affiliations
  • 1New York City Health + Hospitals, New York, New York
  • 2City of Detroit Department of Health and Wellness Promotion, Detroit, Michigan
JAMA. 2016;315(12):1288. doi:10.1001/jama.2015.18035

In Reply Dr Fraser and colleagues and Dr Kiran and colleagues underscore important methodologic issues surrounding J-shaped curves—particularly reverse causality and the need for rigorous research on the precise trajectories of exposure-outcome associations.

We agree with Fraser and colleagues that improvements in causal inference methods are crucial to understanding the public health implications of interventions. However, we caution that instrumental variables may have some limitations in cases in which nonlinear exposure-outcome relationships actually exist. At issue is the assumption that the instrument itself has a linear effect on the exposure variable at all levels of the distribution. Therefore, investigators must ensure that instruments behave similarly at the extremes of the exposure distribution for these tools to be helpful in the setting of J-shaped curves.

First Page Preview View Large
First page PDF preview
First page PDF preview