Pleiotropy-robust MR (PRMR)

Pleiotropy-robust MR estimates the causal effect in the presence of pleiotropy by identifying a patient subgroup for which the genetic variant does not affect the exposure. The pleiotropic effect estimated from this subgroup is then subtracted as an offset from the two-stage least squares (TSLS) estimate. 

Requires a perfect "no-relevance subgroup" and the pleiotropic effect must be constant across covariate subgroups. 

References

• van Kippersluis H, Rietveld CA. Pleiotropy-robust Mendelian randomization. Int J Epidemiol 2017; 47: 1279-1288.

Other terms in 'Pleiotropy-robust one-sample MR methods':

• Adaptive Lasso
• constrained IVs
• MR for gene-environment interactions (MR-GxE)
• MR using G-Estimation under No Interaction with Unmeasured Selection (MR-GENIUS)
• Over-identification tests
• Some Invalid Some Valid Instrumental Variable Estimation (SISVIVE)
• Two-stage hard thresholding (TSHT) with voting