Let’s use some news this morning as an example of how things go in clinical research. You’ll remember that earlier in the epidemic there was a sudden scare about angiotensin-targeting medications (ACE inhibitors and angiotensin receptor blockers, two very common modes of hypertension therapy). This came from the realization that the coronavirus uses ACE2 as an entry point to infect cells, and the worry was that such drugs might upregulate ACE2 protein levels and give the virus an even bigger head start on infecting you.
Since then there’s been a small study from Wuhan that did not show evidence of this effect – in fact, it had a trend towards the opposite effect: patient with hypertension, they found, were at greater risk from the virus (a finding in line with what’s been seen elsewhere), but the ones taking ACE inhibitors or ARBs actually fared better than the ones who weren’t. Today we have a larger study (open access) that seems to confirm that. It’s a retrospective look from the Wuhan area at 1128 coronavirus patients, 188 of them who were taking drugs in those two classes and 940 of them who were not, median age 64, 53% men.
The unadjusted mortality rate between the two groups was certainly different: 9.8% death in the larger group versus 3.7% in the ACE-I/ARB group. This effect persists after a more thorough comparison, adjusting for age, gender, comorbidities and other medications. In fact, it appears that no matter how you slice the data, what subgroups you’re looking at, ACE inhibitors or angiotensin receptor blockers had significant benefit, and that goes for the head-to-head comparison with other hypertension drugs as well.
Now, this is a retrospective study, so it has all the limitations that come with post hoc-ing your way through the data. (I imagine the folks who saw me try to render the plural of “virus” as “virii” will enjoy seeing “post hoc” turned into a verb). The authors mention that an intentional randomized controlled trial will be a much better measure of this effect, and they’re right. But go back to the near panic that people were in when the angiotensin story first hit: if you’d asked folks what they would have expected from a study like this, I don’t think you’d get many takers for the actual results we have. If these drugs do indeed upregulate the amount of ACE2 protein (an open question) then that’s not likely to be making things too much more severe if you can get a reasonably strong signal the other way in a study like this. It may be, in fact, that these results can eventually be taken at face value and that such drugs are actually protective.
Which is, as mentioned, an object lesson in just how much we know about human responses to disease and to drug therapies. One thing I’ve noticed in the comments section here in recent weeks (OK, I’ve noticed a lot of things, but this is a prominent one) are people coming in who take a very straightforward mechanistic view of medicine and drug research. This does this, therefore that does that, and this will obviously raise this and lower those, and we already know that that’s beneficial, QED. Honestly, I wish we could work this way. But we can’t.
After you’ve worked in the biomedical field for a while, you get a lot more cautious about what you know – or you damn well should, anyway. Swinging from paper to paper in Pubmed like Tarzan with his vines in an old movie works in reality about as well as it does if you try the vine trick yourself. You’ll actually end up dangling in the tropical breeze, fifty feet over a pile of sharp rocks. As you listen to your liana unraveling. The drug literature, the biochemistry literature, the medical literature are full of quicksand and crocodiles (to stick with the Tarzan movie metaphors), and it’s a stewpot of some solid results mixed in with the true-but-unextrapolatable, the partly true, the true from a particular angle in favorable lighting, the pretty much false, and no small amount of flat-out gibberish. So confidently proclaiming that hydroxychloroquine opens the zinc channels while azithromycin ripsnipes the frannistan or whatever doesn’t count for much.
It doesn’t count for all that much even if everything you’re saying is reasonable. That ACE2 hypothesis was not crazy at all – perfectly reasonable, one of thousands and thousands of perfectly reasonable hypotheses. We mow those things down like wheat in this business; you get used to it. A good sign these days when you’re listening to someone talk about this epidemic is the number of times they say “We don’t know yet”. The more of those, the better. Confident predictions are grounds for suspicion – under these conditions, the people who are the most confident are probably fools.