[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 79
Citations 0
Comment & Response
November 6, 2019

Heterogeneity Between Genetic Variants as a Proxy for Pleiotropy in Mendelian Randomization—Reply

Author Affiliations
  • 1The George Institute for Global Health, University of Oxford, Oxford, United Kingdom
  • 2Deep Medicine, Oxford Martin School, University of Oxford, United Kingdom
  • 3Collaboration Center of Meta-Analysis Research, Torbat Heydariyeh University of Medical Sciences, Torbat Heydariyeh, Iran
  • 4NIHR Oxford Biomedical Research Centre, Oxford University Hospitals NHS Foundation Trust, Oxford, United Kingdom
JAMA Cardiol. Published online November 6, 2019. doi:https://doi.org/10.1001/jamacardio.2019.4284

In Reply We thank the editors for the opportunity to respond to the methodological issues raised in Gill’s letter and to clarify aspects of our methods. We also thank Gill for his interest in our article1 and for raising important issues in relation to the conduct of mendelian randomization (MR) studies.

We fully agree with Gill in that researchers should aim to use the full range of evolving MR methods to scrutinize the robustness of their findings. More specifically, we agree that the validity of causal estimates from MR studies relies largely on the fundamental assumption of no pleiotropy. As we note in the article,1 our study assumed that the genetic variants selected as instrument for systolic blood pressure affect the outcome only through systolic blood pressure (ie, no pleiotropy). Although we cannot be certain that the genetic variants included in the genetic risk score (GRS) do not have pleiotropic effects, we did not find any evidence in favor of it. While we agree that the MR-Egger test could lead to unreliable results, our interpretation was not based on the MR-Egger test only. The presence of pleiotropy has been checked using individual data from different angles: (1) the association of GRS with possible confounders was assessed; (2) positive control analyses with stroke, heart failure, and coronary heart disease as outcomes were conducted to check the validity of GRS; (3) a negative control analysis with chronic obstructive pulmonary disease was used to further check the validity of instrument; (4) a new GRS excluding single-nucleotide polymorphisms associated with any type of well-known cardiovascular risk factors or diseases was constructed for sensitivity analysis; and (5) patients with coronary heart disease and heart failure were excluded to evaluate the possible mediation effect of these diseases on the association.