Thread by @michaelzlin

First study of Paxlovid in vaxxed people shows a significant benefit in preventing hospitalization for those over 65yo, but not for younger.

The study was independent of Pfizer, used real-world data. Done in Israel of course (US seems incapable of this)

reuters.com/business/healt…

Patients >65yo without prior immunity saw an 86% drop in hospitalizations with Paxlovid. This matches/exceeds the 79% in trials. Those who had prior immunity benefited by 60%, still very good.

I always double-✅ "no effect" findings to see if they're truly neg or there's a possible effect that didn't reach stat sig. The latter can happen due to either small effect size or small study size. For unvaxxed 40-64yo, the risk was 21% (CI 3%–153%) based on 343 hospital cases.

That means Paxlovid may help reduce hosp visits in unvaxxed 40-64yo to 21% vs untreated (a 79% benefit), but to cover 95% of stat possibilities, the estimate is anywhere from 3% (97% benefit) to 153% (53% harm). The wide interval is due to the small # hospitalized at these ages.

Pfizer's Phase 3 trial just grouped all ages together and saw a 79% benefit. The small hospitalization #'s for people <65yo got grouped together with the larger #'s for >65yo. As long as the total p value is <0.05 Pfizer can claim benefit across all ages.
nejm.org/doi/full/10.10…

This reveals one of the ways in which stat analysis in clinical trials is arbitrary. You must choose your analysis groups ahead of time as you're (supposedly) only allowed one roll of the statistical dice.

If you have a small # of patients, you could break out >65yo for largest effect to assure you get p<0.05. But if you have a larger # of patients, you could include younger ages in your analysis and broaden the "finding" and the resulting approved indication.

The predicted effect sizes and p values can be modelled ahead of time. So having resources to recruit more patients helps big pharma get broad indications, but a smaller company who can only recruit fewer patients may have to focus trials on only those patients likely to benefit.

OTOH there's 0 detectable benefit of Paxlovid on 40-64yo with prior immunity. That's zero, as in zip, nada. Relative risk of hospitalization for Paxlovid-treated vs untreated 40-64yo with prior immunity is 118% (95% CI 57%-241%) based on 161 hospital cases.

So there's really not even a hint of benefit of Paxlovid for <65yo. And this Israel study is for high-risk patients (patients with comorbidities such as diabetes, hypertension obesity). That means for low-risk patients <65yo there will almost certainly be no benefit.

So if you're <65yo and have some immunity (which is nearly everyone now), the data reveal that there's no discernable benefit to getting Paxlovid, even if you have preexisting conditions.

Don't expect Pfizer to trumpet these results as it removes the majority of the market.

Here's the study's summary table. The hazard ratios we discussed above. The absolute benefit % take into account the different rates of hospitalizations for the different patient groups. You can see the biggest benefit in reducing hospitalization rates come from unvaxxed >65yo

And here's the link to the study
researchsquare.com/article/rs-170…

Finally to those who complained of COI when I criticized molnupiravir because my lab makes protease inhibitors (like Paxlovid): only crickets from your direction?

We're excited to develop protease inhibitors for those who need it. We don't need to artificially inflate the market

This comes back to my concerns on #Paxlovid overuse by those who didn't need it when test-to-treat was announced

https://twitter.com/michaelzlin/status/1501015509982007297

The ≥12% #rebound rate among people taking Paxlovid recently (essentially all with prior immunity now) is another reason for <65yo to skip it. Otherwise you risk prolonging infectivity period from 1 to ≥2 weeks.

https://twitter.com/santa_care/status/1528063834966982658