Is there any problem on Mendelian Randomization analysis with binary exposure and outcome?

Question

[sjunlee89]

I would like to mendelian randomization (MR) analysis.

In many cases, most of MR analysis has been conducted with continuous exposure and outcome or continuous exposure and binary outcome.

On the other hand, my study design has binary exposure and binary outcome.

For example, binary exposure (=alcohol drinking status (Ever/Never)) and binary exposure (=cancer incidence or not).

However, I could not find the any method for MR analysis with binary exposure and outcome.

ivreg or ivprobit package (STATA or R) can analysis data only with continuous exposure and outcome (both of continuous and binary)

So, what I've wanted know is that there is any statistical problem on MR analysis with bianry exposure and binary outcome.

Even if I've goolged a few days, I could not find how to implement MR analysis in my case.

Please let me know how to go through the problem.

Thank you in advance.

Answer 1

[Amy Mason]

A late reply, but possibly helpful to others - try this paper on MR with binary exposures: https://link.springer.com/article/10.1007/s10654-018-0424-6 and note fig. 1)