OHDSI Home | Forums | Wiki | Github

[Covid-19] Alpha-1 blocker as potential prophylaxis against ARDS

(Aki Nishimura) #1

Hi everyone,

A team of researchers at Johns Hopkins has just initiated a clinical trial of Prazosin as potential prophylaxis against acute respiratory distress syndrome (ARDS) among Covid patients. (This hypothesis has some support from a retrospective study on non-Covid patients: https://doi.org/10.1101/2020.04.02.20051565 .) Due to the large number of Covid-related clinical trials going on, however, the recruitment is slow and it will be another few months before there will be enough data from the trial.

In parallel with the clinical trial, the Hopkins team (consisting of Bert Vogelstein, Chetan Bettegowda, Shibin Zhou, Susan Athey, others) is eagerly seeking evidences from Covid patients data in observational databases. Would there be interests and bandwidths in the OHDSI community to investigate this question? The Hopkins team and I had conversations with @SCYou and @Jaehyeong_Cho, but we probably need a bit more help in actually running the study. I know @msuchard is still working hard at getting a study package ready for Rasinhibitors, but I am thinking that it should not be too much more work to get this study running on the Covid patients data.

Prazosin would a much safer alternative to hydroxychloroquine (HCQ) if indeed effective, so investigating this question could be a good complement to the HCQ study from Study-a-thon.

(Joshua Vogelstein) #2

@aki-nishimura thank you so much for initiating this discussion! And thanks to SCYou and @Jaehyeong_Cho already for your help. We have performed a lot of additional analyses not included in the preprint because they were not ready at the time of submission, including implementing negative controls, multiple methods for propensity weighting, etc. We have also replicated the results in additional cohorts, including pneumonia claims data in the US and elsewhere now too. We would love to be able to ask similar questions using the Korean, or any other COVID data. Happy to share any/all code that we have, etc. Please get in touch if you want to help!


(Mark Shapiro) #3

I’ll ask my team to look in our registry for A1As and get back to you.


(Joshua Vogelstein) #4

@Mark_Shapiro any luck?

(Seng Chan You) #5

@aki-nishimura @jovo I truly apologize that I’ve not been resposive to this thread.
But w’re trucked with small sample problems.
I’ve been thought about Bayesian update to solve this problem. Do you have any idea for this?

(Daniel Prieto-Alhambra) #6

hi @aki-nishimura , a very interesting proposal indeed. I can’t see why we shouldn’t try this! We could see into the feasibility of this in primary care records including potentially CPRD and SIDIAP. Let us chat!

(Mark Shapiro) #7

I sent my team a list of A1A to search for in our data. Will reply as soon as I have something. Trying to get a large enough sample is tough. We’ll be posting our analysis model and the first cut of the dataset this week.

(Aki Nishimura) #8

Thanks for your hard work with the Covid data. I have been working on making an R wrapper for my Python package for large-scale Bayesian regression, but the Python package currently run only on macOS and Linux and it will be a while before it is ready for deployment. In the meantime, I’d be happy to discuss potential short-term solutions to small sample problems.

Great, let me & @jovo know when you would be available to chat. (I am broadly available during normal waking hours in Pacific Time.) We could perhaps have a meeting on Microsoft Teams?

(daniel morales) #9

@aki-nishimura sorry for the late reply too. Discussed with @msuchard earlier in the week. Happy to chat too with @Daniel_Prieto and @SCYou

(Joshua Vogelstein) #10

i’m happy to chat at any time!
please let me know when works.
the original study-a-thon at 3am, 4:30am, etc. eastern standard were possibly sub-optimal times for me, but happy to stay up all night to make some progress on this!

(Cynthia Sung) #11

@jovo Seth Garz at Gates Foundation contacted me (I’m working for Gates Medical Research Institute) to find out whether I knew of any databases we could query to get some info for you & colleagues. In another 1-2 weeks, there’s a Foundation employee whose license to Cerner will become active (PO pending), but he’s quite busy, so not sure he would be able to get to this right away. I am available to join the call. I’m in the Pacific time zone.
As a clinical pharmacologist/PK-ist, I looked at an article on preclinical prazosin experiments (PMID: 29207167) to understand dose- or concentration responses to prazosin and see if it comes close to concentrations achievable in humans at the FDA approved doses (1-5 mg). Vol of distribution is 0.18 L/kg and bioavailability is ~80%. A back of the envelope calculation suggests that the maximum plasma concentration would be <0.08 microgram/mL for the 1 mg dose. The conc in the preclinical studies was 10 mg/mL. The doses used in the mouse study were 10 mg/kg. A very rough rule of thumb is that the human equivalent dose is 1/12 of the mouse dose on a mg/kg basis or ~50 mg for a 60 kg human. There are, of course, many limitations to translating data from in vitro and mice to humans.

(Joshua Vogelstein) #12

@CSung Thanks so much for responding. In our labs’ preclinical work on prazosin (PMID: 30542164), we were regrettably unclear about the precise dosage that we used. After a little investigation, I discovered that we followed the protocols from ref 13 in that work (PMID: 17914358). They used 2 mg/kg, which translates to human dose of 0.16 mg/kg and thus ~9.8 mg per patient of 60 kg.

According to the FDA label for prazosin:

Initial Dose

1 mg two or three times a day (see WARNINGS.)

Maintenance Dose

Dosage may be slowly increased to a total daily dose of 20 mg given in divided doses. The therapeutic dosages most commonly employed have ranged from 6 mg to 15 mg daily given in divided doses. Doses higher than 20 mg usually do not increase efficacy, however a few patients may benefit from further increases up to a daily dose of 40 mg given in divided doses. After initial titration some patients can be maintained adequately on a twice daily dosage regimen.

Our protocols essentially follow this guidance, but only increase to up a total daily dose of 15 mg (target dose, 5 mg q8h) . This is 1.5x the equivalent effective dose in mouse models of lethal cytokine storm.

I regret that we were previously unclear, I hope the above clarifies and addresses your primary concerns?

(Cynthia Sung) #13

@jovo Good to know that you were able to block LPS-induced toxicities in mice at a lower prazosin dose of 2 mg/kg. The interspecies scaling by dose (mg/kg) is a very approximate way to compare dose regimens. Much better would be to have the plasma concentrations from mice (or some other rodent) and humans, by the same route of administration. Do you have any contacts at Pfizer to get the data or someone in your group to comb the web to find the PK studies? Sometimes those are reported in meeting abstracts.

Since prazosin has a half life of 2-3 hrs, minimal accumulation is expected when 5 mg is given every 8 hr for the daily 15 mg dose. Therefore the plasma concentrations are likely to be similar to a single 5 mg dose, though again, plasma concentration data from Pfizer or literature would be more informative. I’m not a doctor, but from the label, it seems that the drug is titrated gradually up to 5 mg/kg q8hr. Would this titration schedule be compatible with the time frame to treat Covid-19 acutely? I also understand that doctors prefer to temporarily stop antihypertensive drugs when patients with acute infections are hospitalized because of the risk of septic shock. Are there any physicians in this Forum who have experience with this drug generally or could comment about practices around use/discontinuations of anti-hypertensives among hospitalized Covid-19 patients?

(Teddy Si Youn) #14

@CSung, @jovo

Dear all, I am very new to OHDSI and a very novice Big Data researcher so have been just lurking for now regarding OHDSI and how to use the data. I hope to be more of a contributor in the data in the near future.

However, I am trained in Neurocritical Care and Stroke clinical care and have worked in Neuro-ICUs for my career. Prazosin as an alpha-1 blocker and fast-acting anti-hypertensive agent is used in that population of patients, but I am not confident to say it is used commonly among medical/surgical ICUs, especially for common bacterial or viral pneumonia patients. In the setting of sepsis, which would be a frequent concern in the COVID-19 pneumonia population, hypotension is a common side effects from infection, sepsis or septic shock. In general, in infectious clinical scenarios, most anti-hypertensives are discontinued “full stop” because of hypotension being a major consequence of sepsis/septic shock. In the setting of severe hypotension, a lot of patients may be put on a pressor, such as epinephrine, while the infection is actively being treated.

I would imaging that low dose prazosin at 0.5-1 mg every 8 hours would have a minimal effect on blood pressure for exactly the half-life that you mentioned in your previous post. Above 2 mg every 8 hours, you will definitely an immediate anti-hypertensive effect that last 2-3 hours per dose.

However, as the patient worsens in their COVID-19 infectious state, I would imagine a few wouldn’t even tolerate the minimal blood pressure drops that would occur with each maintenance dose, although that would have to be trialed clinically.

For the initial post of this thread, I would imagine the study with the use of prazosin would have to be tried early on in the treatment process because too far into it, there would be risk of significant risk of hypotension with continued use. At least, that would intuitively make the most sense clinically.

I hope this is helpful.


-Teddy Youn, MD

(Joshua Vogelstein) #15

@tsyoun Thank you for raising these very important concerns! We fully agree. We are most excited about prazosin and other alpha-blockers as prophylactics. A serious limitation for us getting trials started has been the concern you expressed, as we struggle to convey the idea that we are not proposing a treatment for severe COVID, rather, we are trying to prevent it. We plan to do further pre-clinical studies to investigate the relative timing, and the protocols of the clinical trials state prazosin as early as possible in the disease progression. We are most excited about the outpatient clinical trial, it would be great to be able to keep people out of the hospitals entirely, for their sake, for the sake of the healthcare workers, and everybody else that has other reasons to go to the hospital.

(Cynthia Sung) #16

Thank you Teddy for sharing your clinical perspective! What I love about this community is the diversity and richness of skills and knowledge. Welcome as an active contributor!

(Cynthia Sung) #17

@jovo I didn’t appreciate that you were aiming for a prophylactic treatment. Can you do dose-response and PK sampling in the pre-clinical studies you are planning to inform whether concentrations that match human levels would be protective in the animal models?
Back to observational data, do you think prazosin’s MoA would also apply to patients with seasonal influenza? The OHDSI community more abundant data on that condition.

(Joshua Vogelstein) #18

regarding pre-clinical experiments, we are working on getting permission to get back into the lab to do more, such as the ones you described. also to investigate tamsulosin, which in the retrospective analysis, seems to work, though was never tested pre-clinically.

in terms of influenza, we had a hard time with those data, so much confounding, etc. that is why we focused on pneumonia. perhaps you guys have more experience dealing with such issues, and could do it better, which would be awesome!

(Robert Clark) #19

Thanks for that. By the way, there’s a problem with the link you provided. I think the “.)” got accidentally included with the link.

Reading the article, the researchers had access to over 100,000 cases of patients with pneumonia. With that many cases, using supercomputers, the authors could probably find even unsuspected medications that provide improvement in patient outcomes.

Robert Clark

(Joshua Vogelstein) #20

@Robert_Clark thanks for your interest. here is the link again:

i’m not sure i follow your comment, we demonstrated an effect in two different cohorts, pneumonia (with 100,000 patients) and acute respiratory distress (with 13,000).
in both cases, on multiple outcomes, the effect sizes were quite large, over 50% for the ARD cohort.

the analysis was run on a normal computer, took 20 minutes, and directly tested a hypothesis generated by preclinical data.

we’ve since replicated the results in 4 other cohorts from multiple other datasets, without changing any of the analyses (we are working on updating the manuscript to reflect that).

it feels pretty solid to me.

that said, the reason i am posting on the OHDSI forum is because we are looking for constructive criticism so that we can improve the quality of the evidence.
thus, if you have specific critical feedback on ways we can improve, please do let us know.