Hacker News Reader

Comment by MilnerRoute | Hacker News Reader

Wait a minute... This post is a debate between two opposing viewpoints. So it also makes the exact opposite point:

The primary difference between a randomized controlled trial and meta-analysis is that the former “provide the highest level of evidence because they contain the least amount of bias. Randomized controlled trials reduce bias, while meta-analyses increase bias."

This Hacker News story is about a recent meta-analysis. But an actual randomized control trial of Ivermectin in March (on 476 patients) found the duration of symptoms on treated and untreated patients "was not significantly different."

https://jamanetwork.com/journals/jama/fullarticle/2777389

View on HN

Important thing to know about that study is they are testing time to recovery for people with mild covid.

I'm not anywhere near an expert, but two things stood out to me reading this study. First, time to recover in the ivermectin group was 10 days versus 12 in the placebo. The paper calls this "not significant" but it's not clear to me whether that's statistically not significant, and if so, why wouldn't it be, or if knocking two days off recovery isn't that meaningful. The other thing I noticed reading it was that only one person out of 400 died and that person was in the placebo group.

I think the Brett Weinstein response would be something like -

A. Ivermectin's best benefit is as a prophylactic not a treatment.

B. Ivermectin should be administered as early after onset as possible to treat, whereas here there was some delay to get people organized, enrolled in trial, etc.

C. It's hard to measure effect when the disease is mild.

> why wouldn't it be

Because you only gave the average time of recovery but not the standard deviation.

I didn't see the standard deviation reported in the paper. Is it?

I did see the innerquartile range and it was the same for both the ivermectin group and the placebo. To me that suggests that variation in outcomes is probably similar between the two groups.

They provide it in the form of a hazard ratio:

"hazard ratio for resolution of symptoms, 1.07 [95% CI, 0.87 to 1.32]"

So statistically, those on ivermectin recovered 1.07 times as quickly as those on placebo, but the uncertainty bounds are from 0.87 times as fact to 1.32 times as fast.

> But an actual randomized control trial of Ivermectin in March

Please see this

https://news.ycombinator.com/item?id=27566123

Reply on news.ycombinator.com