2 Comments

Would you please clarify this part?

So they decided that they would subtract one day from the date of the prescription to approximate the time of the positive test. Originally they planned to subtract two days, but apparently for the patients they did have positive covid test data on a bunch of them were getting Paxlovid the same day or one day after, so they moved it up. This was an outpatient study, so they removed any patients given Paxlovid while in the hospital.

Expand full comment
author

Sure, it might be helpful to back up a bit before clarifying. Hopefully this makes sense:

Relevant Question: In patients with covid, did Paxlovid reduce hospitalization?

Control group: Patients with a documented positive covid test in the system. When they tested positive that was day 0 (start watching for outcomes like hospitalization).

Treatment group: Patients given Paxlovid, they hoped to also start measuring outcomes beginning from the positive covid test.

Problem: the majority of patients who received Pax did not have a documented positive test.

Author's solution: In order to approximate that date, they originally planned to subtract two days from the date of the prescription of Paxlovid. They did have some patients with confirmed positive tests in the system given Paxlovid, and they realized most of those patients were getting the prescription either the same day as the positive result or the next day, so they decided to only subtract one day.

My contention: This is a big issue in the context of the other criteria they used. If you want to subtract one day from the treatment arm, adding in that immortal time, it makes the decision to include same-day hospitalization in the control group clearly biased. They included in the control group what looks like at least 30 patients that tested positive and were hospitalized that day. So day zero baseline for control group is elevated. Then by subtracting a day from the date of Pax prescription that guarantees zero treatment patients will be in the hospital at day 0.

This is all ignoring the many important differences between these two groups, so it is just scratching the surface, but clearly the study design was biased to show benefit. One other nugget from one of the supplements is this:

"Note: We are now removing one person who died prior to treatment based on EHR data."

So it looks like they even went through the negative events in the treatment group carefully and extracted a death. The study is full of red flags.

Expand full comment