Mendelian randomization is like the CDOs of the subprime mortgage crisis

Every day there are papers which say x causes y

But the x is not a measured variable

It is hundreds of OTHER variables bundled with weak correlation to X

But because genes are involve many are tricked

Many Mendelian studies provide genetically predicted behaviors

But What’s inside the polygenic risk score for that behavior?

Many papers don’t say.

How does the model perform?

Few papers provide.

Are the genes plausibly associated?

Rarely

At Hepatology Communications we published a study that used Mendelian randomization to examine the impact of a gene on bilirubin levels and debunk multiple spurious associations

That gene literally controls bilirubin metabolism!

journals.lww.com/hepcomm/fullte…