Hacker News Reader

"The researchers analyzed records from more than 1.1 million babies born in Denmark between 1997 and 2016, using the country’s comprehensive medical registries to connect data on births, paternal metformin prescriptions, and birth defects."

Did they look for any defect correlated with any drug? Because if they did, random chance alone would find something.

Anyone read the actual study?

View on HN

I sought it out after seeing this link and found it to be very unimpressive. Science would do well to delete this article spruiking pretty trash science.

P-values seem to be carefully hidden (not in the tables) and as far I could tell were not passing FDR. A lot of models were displayed in various tables and very few of them even had OR CIs that didn't overlap 1.

Hopefully if they didn't apply Bonferroni correction, the study wouldn't pass peer review.

Reply on news.ycombinator.com