Hacker News Reader

Comment by ldoughty | Hacker News Reader

"Lies, ** lies, and statistics"

There was one study that saw 0 participants who contracted HIV during the trial according to the data on the FDA PDF [0]. Was 2,000 participants in Africa who were identified as potentially at risk, aged 16-25.

> YEZTUGO demonstrated superiority with a 100% reduction in the risk of incident HIV-1 infection over TRUVADA (Table 13).

~2,000 given YEZTUGO with 0 infections by the end. ~1,000 given TRUVADA with 16 infections by the end.

Now, this is a great study result if accurate. Substantially better. However, 100% protection is misleading clickbait article. The company does not claim to be 100% effective anywhere I can see... and at best they lifted this statement from this study to use as clickbait.

0: https://www.accessdata.fda.gov/drugsatfda_docs/label/2025/22...

View on HN

Yeah, it's not 100% protection in all studies. One study did have no participants contract aids which is fantastic and would be one data point for 100% prevention.

Another had 2 participants contract HIV out of about 2000 "Person-years". This was compared to another HIV treatment where 9 people contracted HIV (with only 1k "person-years" in that cohort). This equated to 89% reduction in HIV contraction compared to the other PrEP drug.

And that IS a fantastic result and if everyone could take this we'd probably be in a great spot HIV wise. ~90% improvement over current PrEP is great, and it's way easier to take and not mess up.

[1] https://www.askgileadmedical.com/len4prep/understanding/#stu...

What’s a typical rate for infections per person-year among people not using these precautions? For those who don’t know follow the epidemiology here, how good effective are the older drugs compared to not taking them?

Having grown up when AIDS was peaking, the idea of this scourge preventable and treatable feels damn near like sci-fi, and I’m thrilled at the progress we’ve made.

This heavily heavily depends on the population you choose, given the difference in sexual habits.

As a data point, the paper below shows 1,213 out of 18,401 high-risk people in France got infected in 4 years (and 260 out of 31,992 with the previous gen prep, it seems this one reduces it by ~10x again)

https://www.thelancet.com/journals/lanpub/article/PIIS2468-2...

Thanks for that! So yeah, by that, existing PrEP is very effective, and this new one is much better yet.

What a medical miracle, seriously!

I think it's pretty clear that being easier to take and not mess up is the reason for the difference in statistical effectiveness. The reason for lower numbers for effectiveness of daily oral Truvada prep is primarily measuring differences in adherence.

We actually have terms for this.

"Efficacy" is how well something works under ideal conditions.

"Effectiveness" is how well something works in the real world.

So yes - "This is more effective because adherence is easier" is both true and intended.

Just so where clear, from a public health as opposed to basic science standpoint, that's a distinction without a difference.

people magically get more vigilant is as leakly as virus magically goes away on its own.

Yes, my phrasing was responding to the way the parent stated it as "90% more effective, and also it's easier to take". As you say: It's 90% more effective precisely because it's easier, not and also. Behavioral factors matter an enormous amount for the real world success of many types of drugs!

I'd be interested in a modeling study looking at the equilibrium infection rate, assuming everyone was on the drug, but otherwise did not change their behavior with regards to risky sex (or maybe even under a few scenarios of increased risky behavior from risk compensation [0]. You don't actually need 100% protection for the longterm equilibrium to be eradication of HIV (that's the whole idea of herd immunity).

How long would it take for a drug with this level of protection to result in ~no cases of HIV? What level of adoption would it require?

[0] https://en.wikipedia.org/wiki/Risk_compensation

Be sure to model in an anti-vax effect as well

A good use case for the "rule of 3":

>if a certain event did not occur in a sample with n subjects, the interval from 0 to 3/n is a 95% confidence interval for the rate of occurrences in the population.

The cited article is a better source, and it was written a month ago. I am not sure why this is making the rounds now.

They cite 99.9%, and “reduce the risk”, not 100% like this sub article claims.

https://www.gilead.com/news/news-details/2025/yeztugo-lenaca...

Reply on news.ycombinator.com