Patient Attrition Bias Explains The Latest NEJM Ivermectin Study
Patient Attrition Bias Explains The Latest NEJM Ivermectin Study
Their data show the number of people who stayed 100% on protocol. Guess which group didn't stay on protocol?
This:
Is not correct.
The study left out people who failed to adhere to the placebo treatment.
Check this out:
Both groups started with 679 people.
The Ivermectin group had 624 included after attrition.
For the per-protocol analysis, only 288 on placebo were still in the study.
Of the people who adhered to placebo with no benefit, we can expect they would have had more severe COVID-19 symptoms and sought treatment.
Note the patient attrition rate is far less in the Ivermectin group in the “100% Adherence to the assigned regime” analysis, but still biased toward people leaving the study to seek other treatment.
It’s a huge problem if more of one group is dropped from the study due to events that are clearly related to the outcomes measured.
The sickest people in the both groups would have started doing anything other than the pills they were given. The placebo group received visually identical pills in the identical packaging as the active drug, and they left in droves, whereas for the Ivermectin group, they clearly felt well enough to stay on the protocol.
Anyone can see that.
The paper is not to be taken seriously, and should be eliminated from future meta-analyses due to this obvious sign of bias.
Nice find, I knew this study was bunk.. Can you help me understand , is this study the same as the previous one from the TOGETHER trial? It seems there are multiple ones from this same trial, but I might just be confused here..
Hm, I recall finding a possible terminology-driven reason behind the lower "per protocol" number, let me look at the paper again...
*edit: OK, right. The issue was the range of dose durations for the protocol group, not all of which were meant to be compared to the ivermectin 3 day course. "The placebos that were used in the trial involved regimens of 1, 3, 10, or 14 days in duration, according to the various comparator groups in the trial at the time of randomization."
So only the 288 3-day placebo-takers are in the "per protocol analysis" set. And, among the remaining 391, there would be a presumed 1.5% capture to the overall 100% adherence rate for the day 1 group (8.1 x (391/(3x679)), assuming even distribution between 1, 10, and 14) and some type of loss for the longer course groups due to recovery or other (the math would have to weight for days, and would only be a guess anyway). So, the lower adherence could be partially driven by that.