Two-stage approach for causal estimation with time-to-event outcomes, where the second stage regression is substituted with either a Cox proportional hazard, or additive hazard regression.
For this method to provide a valid causal estimate of the exposure on the outcome, genetic variants must satisfy the MR assumptions. In addition, the outcome must be rare for Cox model to be valid, censoring must be independent of mortality and model specification must be correct (i.e., additive hazard, where confounding acts linearly on additive hazard scale). However, there is no consensus about the gold standard way of dealing with time-to-event data within an MR context.
References
- Vansteelandt S, Dukes O, Martinussen T. Survivor bias in Mendelian randomization analysis. Biostatistics 2018; 19: 426-443.
- Tchetgen Tchetgen EJ, Walter S, Vansteelandt S, Martinussen T, Glymour M. Instrumental variable estimation in a survival context. Epidemiology (Cambridge, Mass) 2015; 26: 402-410.
Other terms in 'One-sample MR methods':
- Generalized Method of Moments (GMM) estimator
- Non-parametric methods with bounds of causal effect
- Polygenic risk score approach
- Structural Mean Models (SMMs)
- Two-stage least squares (TSLS)
- Two-stage least squares (TSLS) with binary outcomes
- Two-stage predictor substitution estimators
- Two-stage residual inclusion estimators
- Within-family MR