Quickie
Colonel
No p-values for Hazard Ratio means this is useless, we are unable to determine if the differences are due to random chance or not.
Also, I wouldn't trust TriNetX EHR data, since it's only from 52 organizations in US, it's hardly representative sample size.
Also, propensity score matching leaves unmeasured confounding.
And finally, the conclusion of the article is:
There are confidence levels provided.
(B) 12 week HR (LCL,UCL)
Retinal vascular occlusion 3.54 (3.03, 4.11)
At LCL, HR is 3.03 which is very high. And the high end is 4.11
As to your claim of the study being of a hardly representative size. The N value = 95,156,967 (US Collaborative Network). That is over 95 million subjects.
I have seen so-called authoritative studies in which potentially life-changing decisions were made on the masses that have N-value only in the hundreds or even less.
Last edited: