Covid19Crusher

A larger RCT (n=398), López-Medina, is unfortunately using a lame, composite endpoint, called "Escalation of Care". It combines new hospitalizations or new O2 supplementation>24h or new ICU admission. A sort of weird clinical deterioration synthetic indicator, excluding death.

Post Hoc analysis may be honest and useful, signalling information that was not anticipated in the protocol. However, they are subject to investigator cherry-picking and multiplicity biases and are seen as low quality. Like why did the investigators pick ">12h" and not ">24h"?

Now to the piece de resistance: TOGETHER has selectively disclosed some data on its ivermectin trial, with some very dismissive comments. This is bad scientific form to start with but the urge to badmouth ivermectin was palpable in the words of the principal investigator.

Sadly, we get another lame composite endpoint fusing hospitalizations with "6 hour emergency room observation". This is a much looser definition than hospitalization. ER observation is much "softer": subject to even more random practitioner's discretion. Soft is bad in trials.

Does that allow a doctor who is an ivermectin hater (a common occurrence in Brazil) to bend the data outcome? Sure: put all the participants under observation. Simple. Who's gonna catch you? And ivermectin is toast.

Is the soft "ER observation" a new thing from the TOGETHER trial? Yes, on 2 levels.

A notable point from the López-Medina Post Hoc analysis is that 4/14=29% patients were hospitalized with 12 hours, leaving barely any chance to the drug to make a difference. How many patients in the TOGETHER trial went into observation within hours of randomization?

This could be a material confounder, very applicable to Fluvoxamine as well. It would drag down the performance of the drugs easily.