recent study suggests that post-arrest patients with lactate >12 mM might benefit from TTM33 (as opposed to TTM36). some folks on twitter are getting excited about this, but the study has very weak methodology (#rantorial 1/4)
(study: ncbi.nlm.nih.gov/pubmed/31745738)
(study: ncbi.nlm.nih.gov/pubmed/31745738)
the post-hoc study stratifies post-arrest patients on the basis of lactate level (<7, 7-12, >12). TTM33 vs. TTM36 doesn't make a difference in *any* of three strata. the p-values aren't anywhere close to statistical significance (#rantorial 2/4)
a multivariable model is used to massage the p-value in the lactate >12mm sub-groups from 0.42 down to 0.005. that's an 85-fold decrease! I'm no statistician, but this seems pretty shady (how can you get p=0.005 from a small raw sample size of 4/31 patients?) (#rantorial 3/4)
summary:
(1) start with post-hoc observational data
(2) divide into 3 subgroups
(3) apply multivariable modeling to all 3 subgroups, which drops p-value by 85x in one
(4) reach broad conclusion
read the study & see what you think, but I'm not too impressed (#rantorial 4/4)
(1) start with post-hoc observational data
(2) divide into 3 subgroups
(3) apply multivariable modeling to all 3 subgroups, which drops p-value by 85x in one
(4) reach broad conclusion
read the study & see what you think, but I'm not too impressed (#rantorial 4/4)
• • •
Missing some Tweet in this thread? You can try to
force a refresh
