{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreid6n2iu72l2wgtpdxggtn76lz4lrkzndjk4b6v4mwvaxc7l3wzrye",
"uri": "at://did:plc:wwyqal4cnqhuwyacdj7rqq3n/app.bsky.feed.post/3mepacon3yow2"
},
"path": "/t/statin-related-side-effects-the-recent-lancet-publication-is-biased-toward-false-negatives/28635#post_8",
"publishedAt": "2026-02-11T12:44:36.000Z",
"site": "https://discourse.datamethods.org",
"tags": [
"@ESMD",
"@ESMD"
],
"textContent": "Even though I hate null hypothesis significance testing, I don’t believe that the statistical methods used for analysis of AEs are the explanation for not finding safety concerns, and find @ESMD most convincing. A better Bayesian analysis would convince even more, e.g. for lots of AEs compute P(absolute risk increase > 0.01 or odds ratio > 1.15) scrapping p-values which can detect trivial risk increases.\n\n@ESMD I wonder if you agree with me that reporting counts of AEs in labels shares some of the fault. I’ve always thought that absolute risks should be emphasized much more.",
"title": "Statin-related side effects: the recent Lancet publication is biased toward false-negatives"
}