Skip to main content


Hydroxychloroquine For Avoiding Infection?

Most readers will have heard that this paper appeared in the NEJM late yesterday afternoon: it’s something that we haven’t had so far, an actual randomized, double-blinded, placebo-controlled hydroxychloroquine trial. This one was for post-exposure prophylaxis, a mode of treatment made famous by President Trump when he stated several times that he was taking the drug after people in the White House had tested positive.

It’s from the University of Minnesota, McGill, and the Universities of Manitoba and Alberta, studying 821 asymptomatic subjects, 719 of whom had reported high-risk exposures (either a spouse or other family member testing positive, or health care workers dealing directly with patients). The threshold was enrollment in the trial within three days of exposure, and people already showing symptoms (or a positive PCR test) were excluded from this trial and enrolled in a separate one (which has also concluded, with results coming soon) Participants took either a placebo (407 of them) or hydroxychloroquine (414 people, taking a loading dose of 800mg, then 600 mg 6 to 8 hours later, then 600 mg daily for four more days) and were surveyed regularly out to 14 days, with a follow-up at 4 to 6 weeks. Primary outcome was symptomatic illness, preferably confirmed by PCR, although that wasn’t possible in many instances.

The initial estimate (back in early March) was that 10% of the subjects would in fact develop disease, and the trial was powered to detect a 50% reduction in this number. The team realized that this was basically a guess at that point, and interim analyses were done after 25% and then 50% of the participants had gone through the entire follow-up to re-evaluate. Those revealed first that the infection rate in the control group was running a bit higher (13%), and at the second interim analysis (April 22) the sample size was accordingly allowed to reduce. At the third analysis (May 6) the trial was stopped for futility.

That means what you think it means: there was no statistical difference in infection rate between the HCQ group and the placebo group. On the plus side, no arrhythmias were known to develop in the former (albeit without ECG monitoring, so it would have had to have been severe), and dropout rates were basically identical in the two groups. Side effects were more common in the treatment group, generally GI discomfort. For those who are wondering, there were smaller numbers of people in both groups who were taking zinc supplements, and there was no difference between them, either. At day 14, participants were surveyed to see how well the placebo blinding was holding up (always an interesting measurement). 46.5% taking HCQ correctly identified that they were, 43.9% were unsure, and 10% of them believed that they had been taking the placebo instead. In the actual placebo group, 35.7% correctly thought that’s what they had been taking, 47.6% weren’t sure, and 16.7% believed that they got hydroxychloroquine. Participants who reported any side effects, no matter which group they were in, were 3.7 times as likely to believe that they received the actual hydroxychloroquine, understandably.

So this trial was negative, although one should remember that it could have missed asymptomatic cases. It also tended to enroll relatively healthy people, and can’t speak to any possible protection of high-risk groups. You may also disagree with the initial design to detect 50% reduction in infections, although the HCQ fans would likely have expected at least that much. But within those limits, it provides no evidence that there is a prophylactic effect. From what I can see, the biggest evidence to the contrary is this retrospective analysis from the Indian Council of Medical Research looking at health care workers in that country taking prophylactic doses of HCQ, which found a significant risk reduction. That’s not a controlled trial, of course (and its authors conclude with a statement beginning “Until results of clinical trials for HCQ prophylaxis become available. . .”), but it is worth taking into account. The arguing will continue.

106 comments on “Hydroxychloroquine For Avoiding Infection?”

  1. steve says:

    Before someone else says it, they didn’t include azithromycin, they didn’t include zinc, they didn’t include the kitchen sink. They probably should have tested other latitudes and longitudes. They should have included a whole variety of herbs, homeopathic medicines, energy work and faith healing modalities. They should have done everything under the sun except state the obvious – the stuff doesn’t do squat for COVID and it’s time to stop letting the orange fool in the White House dictate FDA approvals and NIH priorities and spend that money on developing real treatments.

    1. George Stanchev says:

      Yes they did

      “For those asking, 22.4% took #Zinc. Data in the Appendix Table S8. There was no effect of zinc on prevention. 15% with HCQ+Zinc versus 15.3% with placebo+Zinc. Overall, those reporting not taking any zinc had lower (not statistically significant) incidence of symptomatic #COVID”

    2. SaltMineOfStupdity says:

      Yeah but did they try Zinc Phosphate? :- :

      1. steve says:

        Zinc gluconate? Quick, buy ColdEze stock!!!

      2. philip alabi says:

        Funny. Lol

  2. S Silverstein MD says:

    It’s good to see what appears to be a relatively robust study.

    Regarding the company Surgisphere – how can they plausibly have obtained so much HCQ data without the hospitals from which the data allegedly came having known?

    Here’s something I wrote a decade ago that offers a potential answer:

    “Health IT Vendors Trafficking in Patient Data? ”

    1. Tom says:

      Or, they could just make it up.

  3. Is it Fake News like the Lancet paper that you praised?
    Lancet chloroquine scandal.
    – No reproducibility
    – No data traceability
    – No review transparency

    1. David Young MD says:

      No, sorry. This one was a well designed randomized clinical study. Small one, perhaps, but still a randomized study.

    2. Derek Lowe says:

      I try not to throw around the term “fake news”, TBH. Don’t like its provenance.

  4. Dr.Puli says:

    The beauty of this double-blind placebo-controlled study evaluating the prophylactic application of hydroxychloroquine is muddied by the fact that the final outcome ( who got the infection) was ascertained by RT-PCR in merely ~ 20%, rest were based on symptoms ( Ooops), which is highly misleading and open to assumptions. Why in the USA there was a shortage of kits, and this one destroyed the rigor of this otherwise beautiful study. Also, sad to see political affiliations becoming more and more explicit in scientific conversations that are supposed to be unbiased. Regardless, HCQ deserves fair treatment.

    1. Nickname says:

      Are we really surprised that this is so politicized given the administration’s politicization of everything from the health benefits of asbestos to our “clean beautiful coal”?


    2. Jocco Dundee says:

      It is shameful/disgraceful that so much time, money, energy, and hope is placed in a trial that turns out to be designed by incompetent “scientists”. How is it possible that you could run a trial to test for the prophylactic antiviral effect of a drug, and not test EVERY SUBJECT using the well established viral diagnostic (RT-PCR) at baseline and then at regular intervals throughout the trial?

      This trial could have been so informative, and laid to rest the HCQ story for prophylaxis once and for all. Instead, it just adds to the uncertainty. It would have been better not to run the trial. I got two RT-PCR COVID tests in the last week ($65 each), both negative thankfully. There is no reason given the resources invested in this trial that they would not have been able to include COVID RT-PCR tests throughout. Pharmacokinetic and biomarker samples are routinely included in large trials for every subject,, and those are orders of magnitude more difficult to validate and implement.

      Can someone help me understand why so many of these trials are so poorly designed? Is it fear and anxiety of the pandemic that is causing scientists to lose their minds? The recent Remdesivir trial also did not publish viral loads (they may have measured it, but it wasn’t in the publication). Is there any possible mechanism of action for Remdesivir that does not require reduction in viral load?

      1. Robert Clark says:

        I agree with you it is odd that so many bad studies are being run about COVID-19. There are a few other early treatment studies being run on HCQ. Let’s hope they are designed better than the ones so far.

        Robert Clark

      2. Another Guy says:

        The only perfect clinical trials are ones that have correctly identified all possible variables and controlled for them and essentially know the outcome a-priori. In other words, the perfect study is the one that doesn’t need to happen. Example: 45 magnum gunshot to center of head at point-blank range and survival rate.

      3. theasdgamer says:

        Look at when the study began and tell me about PCR tests at that time.

      4. Fred says:

        poorly designed studies….. maybe the people designing the studies are afraid of the results, how they will be perceived, and want to leave themselves wiggle room. the environment has become so political that no one wants to be the one who published the study that confirms that Trump was right.

    3. Robert Clark says:

      Yes, most subjects didn’t have a COVID-19 test at the end of the Boulware study. Most of the end results were just by subjects self-reporting of symptoms. But as Mr. Lowe noted it is well known that most COVID-19 infections are asymptomatic. It also seems plausible that in many cases people will think they have symptoms of COVID-19, when it is actually something else. This is borne out by an Indian study that did show HCQ has benefits towards preventing COVID-19:

      The study found only 5% of symptomatic people tested positive for COVID-19.

      BTW, the web address to the Indian study given there expired. You can find the study saved here:

      Healthcare workers & SARS-CoV-2 infection in India: A case-control investigation in the time of COVID-19.

      Because of the necessarily large number of both false positives and false negatives in the Boulware study, little conclusions can be drawn from it.

      Robert Clark

      1. Edward R says:

        Dr. David Boulware forgot to disclos his previous Gilead research grants in my opinion. (Very bottom)

        Since few of his subjects were actually tested for CoVid-19 and such a high dose (1400mg first 8 hours) would be likely to induce common (1-10%) HCQ side effects of diarrhea and headaches which would then trigger a CONFIRMED CV-19 result in his study… well… it is entirely possible that NO CV-19 infections occured in the 400 patient self reporting treatment branch.

        In any event the HCQ arm still reported lower HCQ/CV-19 symptoms.

        Junk Science at it’s worse.

    4. theasdgamer says:

      Symptoms…didn’t even mention pO2 levels…didn’t understand the disease. Maybe mistook flu for covid?

  5. RA says:

    FYI to those who take their medical cues from Dr. Trump…when he used HCQ for post-exposure prophylaxis, he had ECG monitoring…

    “The drug was administered “in consultation with [Trump’s] appropriate care team members and close monitoring of the electrocardiogram” for irregularities, Conley wrote.”


  6. brian says:

    Well, this actually doesn’t “prove a negative.” Given that the point estimate was different, what it does show is that the study was powered to detect a certain clinical effect, and the study failed to detect an effect at that level. It DOES NOT “prove” that Hydroxychloroquine has NO clinical effect. If you think it does, Derek, then I’ve lost a lot of respect for you this morning.

    If this had been an Alzheimers disease study, and you saw a point estimate incidence of people developing the disease over the follow-up period of 11.8% versus 14.3%, that would be an amazing result and people would be falling all over themselves to buy the stock of the company which made the drug. But, then again, they probably would have enrolled more than 800 people in that study, so it would also have been a statistically significant result.

    A 20% reduction in disease incidence in either COVID-19 or Alzheimers would not be a terrible result. That would be a very good result. And COVID-19 is infectious, so a reduction in incidence also prevents other infections, hence the overall impact would be higher if “everyone” was taking it.

    Since the study wasn’t powered to detect that level of effect, it is just a point estimate and not statistically significant. But it is certainly NOT proof of lack of effect, either.

    I think too many people have let their political beliefs get in the way of their scientific judgement. Please, let’s keep politics out of this discussion, because it simply too important to get it right.

    1. Chris Dockendorff says:

      You are correct Brian that “It DOES NOT “prove” that Hydroxychloroquine has NO clinical effect”. But what is your point? That is true for EVERY clinical trial that does not prove a statistically significant effect. The objective of these rapid trials is not to prove Dr. Trump incorrect, it is essentially to determine how we should invest our precious healthcare resources– more HCQ studies, or greater investment in more promising alternatives. Your comparison to AD is correct but misguided– absolutely nothing has yet shown in the clinic any promising effects, after decades of effort. We are just beginning to tackle COVID, and we already have a far better understanding of its etiology than AD. Every day and every dollar we waste on an unpromising therapy is a day and dollar less we have for something that could make a real difference for patients in urgent need.

    2. anon says:

      Can’t you read? “You may also disagree with the initial design to detect 50% reduction in infections, although the HCQ fans would likely have expected at least that much. But within those limits, it provides no evidence that there is a prophylactic effect.”

      1. David Young MD says:

        Exactly! If you asked a rabid pro-hydroxychloroquine person how well would HCQ work to prevent infection, they would probably tell you that HCQ would prevent everyone from getting an infection. That’s the hype. So the fact that 12 percent got Covid19 with HCQ is a dismal failure for the drug (in light of 14 percent who did not take HCQ). The results are not at all what the pro-HCQ people would have expected.

        1. Robert Clark says:

          Unfortunately, you can’t draw that conclusion from this study due to poor initial design. This is because less than 20% of the subjects were actually tested at the end. Most subject results were just by self-reported symptoms.

          It’s already well known 80% to 90% of infected people are asymptomatic. It’s also true that apparently “symptomatic” people really don’t have it, to the tune of 80% as well.

          When you have possibly 80% of the results being wrong, you really can’t draw any conclusions either way.

          Robert Clark

          1. theasdgamer says:

            If you limit the symptom definition to pO2<94 & adyspnic, then you have a moderate covid case–and flu can be ruled out.

        2. Neil Ferguson says:

          Let’s put your rather malicious characterization of proponents aside, MD. Is this alternate statement correct? “The study showed only a modest reduction of infection rate (15% lower than placebo). Unfortunately the results are within the statistical margin error for the sample size. So actual benefit could very well be 0, or it coulbe be substantially higher than 15%. Still, not the major reduction that everyone (except partisan sociopaths) was hoping for.”

          1. theasdgamer says:

            I answered your question about prophylactic significance near the bottom.

    3. Jocco Dundee says:

      On the comparison with Alzheimer’s disease, imagine running a clinical trial of a gamma-secretase or beta-secretase inhibitor and NOT measuring change in amyloid or abeta in every subject. That would be ridiculous and incompetent.

      What we really need here is a properly designed dose-response trial that evaluates prophylaxis via viral load measures with RT-PCR (not flu symptoms!). If none of the doses of HCQ (or whatever antiviral drug) show any statistically significant improvement in either positive cases or viral dynamic time course, both evaluated by RT-PCR in ALL subjects, then do not even bother conducting a larger clinical outcomes trial.

    4. Ian Malone says:

      Having followed this blog for a long time, very very many of its readers *would* have been highly sceptical of the value of a 25% reduction in AD progression rate (see the aducanumab articles for most recent examples). We have seen claims this particular drug is a ‘game changer’, that it’s a ‘cure’, then that it wasn’t being tried early enough, and now the claim is that studies aren’t powered enough to detect its slight effect. If it has one it’s beginning to look very like it’s small, as has been expected for almost any re-tasked existing drug in this fight.

      What would a 10% reduction worth? Is it enough to offset the safety concerns for talking a drug prophylactically? Is it more than the benefit of wearing masks or maintaining social distancing? And that’s still only a hypothetical, because the study not being powered to detect a small effect is not the same as the effect being there.

      Maybe 10% would be worth it to those who couldn’t avoid exposure, like healthcare workers (again, simpler to focus on giving them proper protective gear), but right now the existence of any effect remains an article of faith.

  7. luysii says:

    What would have been really interesting would have been antibody testing before and after prophylaxis. We know the antibody positivity rate is an order of magnitude higher than the PCR rate, reaching an astronomical 33% in the Bronx, proving that most infections are asymptomatic.

    1. john says:

      Theoretically, Boulware et al could still do an antibody study of the patients whom they judged to be positive for symptoms, but in whom PCR was not done, or, for that matter, in whom PCR was done but was negative despite symptoms.

      Of course, not all qualifying patients would agree, but it would be a good ancillary study that could (and should, IMO), be done.

      BTW, here is an interesting study published in Annals of Internal Medicine (from Boulware’s twitter feed) re PCR positivity in COVID-19 patients. Apparently the lowest false negative rate is on the order of 20%.

      This means that AT BEST, the PCR test is MISSING 20% of COVID-19 cases. Of course, this depends on the pre-test probability of COVID-19.

      1. JasonP says:

        New Test approved for COVID-19 total anti-body. Better sensitivity & specificity, 100% / 99.8%

        >>>> Siemens Healthineers announced today the U.S. Food and Drug Administration (FDA) has issued an Emergency Use Authorization (EUA) for its laboratory-based total antibody test to detect the presence of SARS-CoV-2 antibodies including IgM and IgG in blood. Test data demonstrated ******100 percent sensitivity2 and 99.8 percent specificity.****** The total antibody test allows for identification of patients who have developed an adaptive immune response, which indicates recent infection or prior exposure. Testing can begin immediately with more than one million tests already shipped to health systems and laboratories.<<<>>The company is prepared to ramp up production as the pandemic evolves with capacity exceeding 50 million tests per month across its platforms starting in June.<<<

        So by December we should be able to test 90% of the USA?

        1. Some idiot says:

          Hmmm… interesting!!! 🙂 Any idea how that converts to false positive% and false negative% …?

          1. Sensitivity = probability of true case being detected, so false negative rate = (1 – sensitivity)
            specificity = probability of detected case being true, so false positive rate = (1 – specificity)

          2. luysii says:

            It’s tricky and all depends on what the actual rate of antibody positivity is in the population. With a 1% rate and a test with 99% sensitivity and specificity, half the positive cases will be false, and half of the true cases will be missed. Now consider the Bronx with a 33% positivity rate for antibodies (up to 44% in some neighborhoods). Almost all the true cases will be found with the test and while the number of false positives is nearly same, as a % of the true rate is it small.

  8. ol' bonespurs says:

    orange menace white house
    snorts hydroxychloroquine
    twitters excrement

    1. Peon says:

      Stupid, I know, but that made me laugh out loud.

  9. David Young MD says:

    There are plenty of rabid pro-hydroxychloroquine people who say “12 percent rather than 14 percent?… why you can extrapolate that to 10,000 people and you can see that you prevent 200 infections!” It is this sort of craziness that abounds out there. Don’t worry, there are another 10 studies underway testing HCQ as a preventative. There will be all sorts of results, put they will regress to the mean of showing the control group and the treatment group to be the same… I think.

    My retort to all of those Youtube people who post and say that 12 percent is better than 14 percent, “Hurrah!” is…. “Tell me, before this study was started, what different in outcome would you have guessed? Probably a 20 to 30 percent difference, I would think.

    Then again… you just can’t convince them. Once you believe that there is a conspiracy……….

    But maybe you can convince a casual reader.

    1. Sanjay Kaul says:

      If the trial had been designed to detect a more plausible effect size, say 25% reduction in symptomatic illness, the sample size would quadruple (related to inverse square law which states that halving the effect size results in quadrupling of the sample size) making the trial not feasible. On the other hand, inadequately powered trials amplify the potential of a false-negative result (and wasting precious resources). Reconciling trial feasibility and clinical relevance is always challenging!

      1. Charles H. says:

        In addition, while it wasn’t powered to detect a 25% effect, one would expect at least a not-statistically-significant indication in that direction if the effect existed. That is wasn’t there doesn’t rise to the level of proof, but it *is* a strong indication.

        The lack of tests is more problematic, but given the problems with testing during that time period, it’s quite reasonable. Just extremely annoying.

    2. Robert Clark says:

      From the oddness of the Surgisphere affair, I wouldn’t even count out a conspiracy theory as having some merit at this point.

      BTW, my background is in science, but I’m interested not only in the science but also in the history of science. Something I noticed occurs in science is that something that should have been obvious is missed even by the leading experts in a field. It sometimes takes years to decades for this key observation to be made, when it should have been obvious years earlier.

      There are a few obvious things being missed in regards to COVID-19 that are real head scratchers. For me the top thing is, experts in epidemiology and infectious disease seem to have no qualms in HCQ being judged solely for patients under severe disease. It’s like textbook, course no. 101 that antivirals are most effective when given early before severe disease progression sets in. Yet experts in the field take studies that show, for example, HCQ ineffective in severe disease prove that it is ineffective in early use. It is well known one does not prove or disprove the other but you have experts in the field talking like it is the same thing.

      Then there’s the Surgisphere affair that should have raised red flags up and down the line for experts in the field.

      Finally, of the obvious oversights of the ones I noticed is this study by Boulware Even by when they started recruiting participants in mid-March it must surely have been known that most COVID-19 infected people are asymptomatic. Then to determine whether someone is infected or not by self-reporting of symptoms should have been seen from the beginning as highly inaccurate. And it turns out that it’s even worse than that since you can also have people reporting “symptoms” when they really are not infected.

      Puzzling cases of the “obvious” being missed in science.

      Robert Clark

  10. Bob says:

    Question for clarification: “For those who are wondering, there were smaller numbers of people in both groups who were taking zinc supplements, and there was no difference between them, either.”

    What does “no difference between *them*” refer to here? Between those taking zinc in the HCQ group and those taking zinc in the control group? Or between those taking zinc and those *not* taking zinc, regardless of group?

    1. Derek Lowe says:

      Any way you look at it, actually – have a look at the Supplementary Files for the paper and you can see for yourself.

  11. john says:

    So, here are my thoughts on the Boulware post-exposure trial.

    In addition to the problems already discussed (low-event rate, low-percentage of symptom confirmation by PCR, study done at the tail-end of the flu season),
    there are more basic issues.

    The most important issue I believe is, did the study ask the right question?

    I remember in interviews with Zelenko, he was asked about whom he gave HCQ to, and what the results were. So, I’m paraphrasing here. I think this was an interview he had given to a guy called Corsi, but not 100% sure.

    Q: Do you treat everyone with HCQ?

    Z: No, only those who are over 60, or younger persons who have risk factors.

    Q: And so, when you give HCQ, do the patients have no symptoms?

    Z: No, they still have symptoms. Just milder.

    The outcome of this trial was absence of COVID-like symptoms. This is an outcome that no one, not even the advocates of HCQ, were claiming.

    One can understand this better by comparing HCQ vs. COVID with oseltamivir (Tamiflu) vs. influenza.

    Tamiflu has been shown to be effective agains the flu in a number of randomized trials and meta-analyses. However, (and I haven’t gone over all of the Tamiflu studies with a fine-tooth comb), the benefit of taking Tamiflu is never (or almost never) an absence of flu-like symptoms, but rather, a reduction in severity and/or duration of symptoms. In the latter case, usually a reduction by 1-2 days.

    In the Boulware NEJM study, there was no prespecified analysis of symptom severity.

    Recently, I was concerned about risedronate causing flu-like symptoms. Risedronate has been reported to cause such symptoms. The best way to look at such questions is to look at placebo-controlled trials, and compare the incidence of symptoms in the drug group vs. the placebo group. In one study I quickly found, the incidence of flu-like symptoms (of course, over a longer time frame than in the COVID study) was 10% in both the risedronate group and the control group.

    So, in way, Boulware et al were asking a question that no one else was asking regarding a potential benefit of risedronate.

    Now, with a much larger event rate, one could argue that with milder symptoms, and incidence of symptoms should also be reduced slightly, and maybe it was in this case (12 vs. 15% being 25% reduction). But Zelenko never claimed that taking HCQ would totally block COVID symptoms, and Raoult never advocated for HCQ (to my knowledge) in any type of prevention role.

    1. john says:

      Well, maybe I was not 100% correct in the previous post. I found this paper in JAMA by Welliver in 2001 re oseltamivir and post-exposure prophylaxis. Oseltamivir reduced household transmission by 89%. So, the effect of HCQ in the Boulware trial, if present looking like around 20-25% (which would require a much larger sample size, say 2000-3000 patients to confirm), is underwhelming.

      Effectiveness of Oseltamivir in Preventing Influenza in Household ContactsA Randomized Controlled Trial

      Robert Welliver, MD; Arnold S. Monto, MD; Otmar Carewicz, MD; et al Edwig Schatteman, MD; Michael Hassman, DO; James Hedrick, MD; Helen C. Jackson, PhD; Les Huson, PhD; Penelope Ward, MD; John S. Oxford, PhD; for the Oseltamivir Post Exposure Prophylaxis Investigator Group

      JAMA. 2001;285(6):748-754. doi:10.1001/jama.285.6.748

      Context Influenza virus is easily spread among the household contacts of an infected person, and prevention of influenza in household contacts can control spread of influenza in the community.

      Objective To investigate the efficacy of oseltamivir in preventing spread of influenza to household contacts of influenza-infected index cases (ICs).

      Design and Setting Randomized, double-blind, placebo-controlled study conducted at 76 centers in North America and Europe during the winter of 1998-1999.

      Participants Three hundred seventy-seven ICs, 163 (43%) of whom had laboratory-confirmed influenza infection, and 955 household contacts (aged ≥12 years) of all ICs (415 contacts of influenza-positive ICs).

      Interventions Household contacts were randomly assigned by household cluster to take 75 mg of oseltamivir (n = 493) or placebo (n = 462) once daily for 7 days within 48 hours of symptom onset in the IC. The ICs did not receive antiviral treatment.

      Main Outcome Measure Clinical influenza in contacts of influenza-positive ICs, confirmed in a laboratory by detection of virus shedding in nose and throat swabs or a 4-fold or greater increase in influenza-specific serum antibody titer between baseline and convalescent serum samples.

      Results In contacts of an influenza-positive IC, the overall protective efficacy of oseltamivir against clinical influenza was 89% for individuals (95% confidence interval [CI], 67%-97%; P<.001) and 84% for households (95% CI, 49%-95%; P<.001). In contacts of all ICs, oseltamivir also significantly reduced incidence of clinical influenza, with 89% protective efficacy (95% CI, 71%-96%; P<.001). Viral shedding was inhibited in contacts taking oseltamivir, with 84% protective efficacy (95% CI, 57%-95%; P<.001). All virus isolates from oseltamivir recipients retained sensitivity to the active metabolite. Oseltamivir was well tolerated; gastrointestinal tract effects were reported with similar frequency in oseltamivir (9.3%) and placebo (7.2%) recipients.

      Conclusion In our sample, postexposure prophylaxis with oseltamivir, 75 mg once daily for 7 days, protected close contacts of influenza-infected persons against influenza illness, prevented outbreaks within households, and was well tolerated.

  12. Simon Auclair the Great and Terrible says:

    No, Derek, noooooooo!

    1. sunyilo says:

      Jocco Dundee,
      In hindsight everything is 20/20. Don’t forget these studies were designed in early March when one could not get a test wherever and whenever patients needed. The authors deserve credit that they are fully transparent about their methods and clinical data (like the definition of a good bike: you can take it apart and put it together easily without breaking it). Under such pressing global medical needs you don’t have the luxury to have time to cross each and every box on your trial checklist. Sure there are better ways to conduct such trials during the time of the worst pandemic of the last 100 years, but that, unfortunately needs politicians to put aside their ego and ideology and facilitate international efforts without second thoughts.

  13. thx zetec says:

    Derek – I look forward to you articles and appreciate you expertise and communication skill. Also your objectivity – you published piece criticizing the Lancet HCH-heart-risk study.

    I am little confused by what this latest study did w/ zinc. It was said there was zinc involved somehow, but when I click on your link I don’t see no zink (sorry couldn’t resist). The link to study seems to show all text and tables, and detailed list of procedures. I did text search “zinc” maybe it is in a graphic table and I missed it.

    1. john says:

      The zinc data are in the supplementary materials file. You need to download that separately.

  14. RA says:

    So, I agree the study isn’t definitive for many reasons, but so far little good evidence of effectiveness. We’ll see what further studies show.

    But to the conspiracy theorists, I am curious why you think Donald Trump hasn’t used his power to get a big, methodologically rigorous trial testing the Zelenko protocol so far that you would accept? I mean, the guy will use his power to gas peaceful protestors for a photo op…so why won’t he lift a finger to get a decent trial that would satisfy you all? It’s Operation Snail’s Pace for HCQ, not Operation Warp Speed.

    On some level, I admire the tenacity of the pro-HCQ foot soldiers and maybe the data will one day show something, likely modest, to your cause…but your “general” is hiding in the bunker on this issue, but not actually making a study you would accept happen anytime fast!!!

  15. KeepinItReal says:


    Another flawed study masquerading as the last word.

    Poor adherence to the supposed regimen of HCQ+Zn+antiviral.

    Poor endpoint choice.

    How hard is it to do the right experiment?

    1. Sephirakra says:

      You’ve never worked in clinical research, have you?

  16. David says:

    “the trial was stopped for futility. That means what you think it means: there was no statistical difference in infection rate between the HCQ group and the placebo group.”

    Not what that means. Futility means the trial was stopped because, if continued, it had no reasonable chance of ever showing a difference. If you look at the full paper, you’ll see some statistical mumbo-jumbo about spending functions, which are methods of correcting for multiple testings and avoiding false results due to multiplicity, and a statement that “the trial was halted on the basis of a conditional power of less than 1%, since it was deemed futile to continue.”

    1. john says:

      I think that the study was powered to detect a 50% difference in symptoms, which is fairly huge. So, the interim analysis would have suggested that it was pointless to go on, as a 50% drop in incidence of COVID symptoms was very unlikely to be detected with further enrollment, given only a 20% or so trend favoring HCQ with ~800 patients enrolled. It was decided ahead of time that a 50% drop in incidence was what would be clinically important and that’s what the study was designed to detect.

      If an outcome of a 20-25% drop in incidence of symptoms would have been the goal, then the study would have continued. But with an event rate of only 10-15%, the study would be required to recruit more than 2000 patients to be able to get sufficient power to detect or exclude a 20-25% benefit.

      With sample size calculations, it’s a bit of a game. As the percent improvement that you want to detect drops, esp. down below 20% or so, then the number of patients required to exclude a false negative rate rises exponentially. The other variable is the event rate. If the event rate is high, say 50% or so, then you don’t need a lot of patients to detect even relatively modest improvemenet. But if the event rate is in the 10-15% range, then you need tons of patients to detect a modest improvement.

      Maybe there are some statisticians on the blog who could give more info on this, but this is my understanding of things.

  17. John Wayne says:

    Okay, here is the study I want to do:
    1. Two cohorts, hydroxychloroquine and placebo
    2. Each cohort divided by political affiliation (self-identified moderates are excluded)
    3. Tell the cohorts getting the real drug what drug they are getting

    This is pretty much a sociology experiment, but I’d love to see the placebo effect play out here.

    1. JasonP says:

      Doesn’t the placebo effect have the potential to account for a 20% “response?”

  18. cynical1 says:

    I am going to ask a scientific question regarding drug pharmacokinetics not related to this trial….. I noted that the trial design above was patients were dosed for 5 days and then followed for 14 days. (I thought that the pro-HCQ crowd would immediately start chanting, “They didn’t take it for 14 days!!! Bogus!”. Y’all disappoint me but feel free to pick up this new mantra.)

    Anyway, that’s not my question. I also remembered seeing a really long half life for HCQ in the past…….and it is. It’s about 22 days with about 74% bioavailability. (Probably an amazing drug for malaria prophylaxis.) But I now just looked up dosage in lupus patients and these patients are taking 200-400 mg daily. How in the world does this drug not accumulate to the point of its LD100 over time in these patients? Are there p450s that get upregulated with accumulation so that it reaches some safe steady state level and stops accumulating? Otherwise, with that half life, I don’t see how it ever stops accumulating in vivo? I didn’t see anything that suggested that it wasn’t being dosed chronically in lupus patients either. What am I missing here?

    1. Mike K says:

      Because the clearance of a drug is proportional to its concentration the amount accumulated doesn’t necessarily go up forever. Eventually you get a systemic concentration so high that the rate of elimination equals the dosing rate and you have steady state. You can estimate how long it takes to get to steady state by multiplying the half-life by 5. That’s not to say that the exposure at that point (about 100 days for HCQ) won’t be pretty high.

      1. cynical1 says:

        Thanks for your answer, Mike K.! It makes sense now. Now that you mention it, I do vaguely remember that part about 5 half lifes. (It’s been many, many, many years since I took my PK class.) With that said, after 100 days, I’m surprised the stuff isn’t precipitating out in your blood stream. You could probably use NMR to quantitate the AUC at that point. It also suggests to me that when it comes to COVID-19, the biggest question with HCQ is efficacy and less with toxicity.

        1. NICK says:

          And 7.5 x 1/2 life for “complete” elimination.

  19. anon21 says:

    Dozens of scientists called into question the recent lancet study that the WHO used to stop the HCQ treatment, and now the WHO resumed it (data was falsified). ……But, I thought the “science was settled”???? Wasn’t it supposed to be just trump-supporting lunatics that still thought HCQ treatment had a prayer?

  20. georgi momekov says:

    The Lancet fraudulent study has been retracted…
    Sadly, the credibility of all kind of research will be questionable if such fraudulent studies continue to be published, probably just because one of the co-authors is a professor at the Harvard Medical School…

  21. Bunker Bitch says:

    Yum, more hydroxyklorox!

  22. 11 fingerssss says:

    I’ll drink to that.

  23. theasdgamer says:

    Could someone please look at this for me who has access? They used PCR testing in their study.

    1. Robert Clark says:

      It was saved here:

      Healthcare workers & SARS-CoV-2 infection in India: A case-control investigation in the time of COVID-19.

      Robert Clark

      1. theasdgamer says:

        Thank you.

  24. WustlMed says:

    Another bogus publication. The authors made a wrong claim. If the trial was to test the 50% reduction, the conclusion should be something like “we failed to prove the 50% reduction, the data suggest that there is only a 20% reduction…”.

    1. Druid says:

      Unfortunately, that is not how statistical science works. The study was not big enough to have enough power to test a smaller difference than 50% with the likelihood (eg 95%) that is expected. Yes, there was a 20% difference, but the random chance of getting a 20% difference between the two groups was too high. In fact, in the summary, the authors say the -2.4% difference (11.8% vs 14.3%) had the 95% confidence interval from -7.0% to +2.2%. That is to say that no difference – 0% – cannot be excluded at the 95% level. The final sample sizes – 49 and 58 – are too small. This is disappointing, of course. This information could help you design the size of study capable of testing a 20% improvement, but you cannot claim that it supports that level of improvement. Sorry.

      1. WustlMed says:

        “suggest” != “support”

        1. Druid says:

          Sorry, I did unintentionally exaggerate your comment from “suggest” to “support”. No study is conclusive. You could equally claim that the data “suggest” HCQ makes no difference at all because 0% falls within the confidence interval. I doubt if there is sufficient power to “support” this claim either.

          1. WustlMed says:

            Have to be loyal to the data. The data gave a point estimation of about 20% reduction although with a large confidence interval.

          2. Druid says:

            WustIMed – I think that is where you are going wrong. The (modest) difference was seen in a small sample. There is no reason to believe that it will be reproduced in the population or even in a bigger sample.

          3. WustlMed says:

            In fact, I don’t believe that, and we shouldn’t. It might be 0% or 40% (need to check the confidence interval). And this is why the author did wrong when they claim “no effect”. “large confidence interval” != “no effect”.

          4. Druid says:

            Where did they write “no effect”? I can’t find those words, and I have searched the paper. All I can see is “The incidence of new illness compatible with Covid-19 did not differ significantly …”. I would say they were very careful to write only claims supported by the study design, assumptions and data.

      2. Another Guy says:

        I agree, to design a good study, one must first have a very good idea of the expected difference between the two study arms, and then you can calculate how many patients you need to enroll for the statistics to work. How do you do that? You look at the literature and see the results of previous studies that tested the same drugs on the same population with the same disease. What happens when you have a new-to-humanity viral disease? You guess the expected differences and cross your fingers. Does that mean science is broken? No. You use the data you gathered from the first experiment to design better experiments. What is a poor-quality clinical study? That would be an experiment that doesn’t help you design the next study, so you risk repeating the same mistakes. We’re seeing a lot of poor-quality COVID-19 studies these days because of the urgent need to get started on something when many of the variables are unknown or difficult to control under the circumstances. However if there is one nugget of truth in it, then science can progress. What is a really bad study? The ones that are clearly fraudulent and intend to mislead, either by data-fakery or deliberately skewing the design to favor a pet theory rather than put the theory to the test.

        1. WustlMed says:

          “After high-risk or moderate-risk exposure to Covid-19, hydroxychloroquine did not prevent illness compatible with Covid-19 or confirmed infection when used as postexposure prophylaxis within 4 days after exposure.”. How blind to their data the authors had to be to claim this.

  25. Bob says:

    So what this study tells us is that HCQ cannot uninfect better than placebo people who have been infected for 1 to 4 days already and eventually show symptoms of CV19, but are not actually confirmed to have CV19 by test. Prophylaxis on patients who are already infected for several days? It was an excellent opportunity to evaluate early treatment and instead the end point was did they have symptoms of CV19?

    FWIW, check out Day 1 HCQ

    Symptomatic Infection Rates by Days Exposed before treatment
    Day ——–> non-HCQ——>HCQ
    1 ———-> 12.7 ————-> 6.5
    2 ———-> 17.0 ———-> 12.0
    3 ———-> 14.5 ———-> 12.2
    4 ———-> 12.4 ———-> 14.5

  26. theasdgamer says:

    The docs didn’t even look at pO2 values. Very weird. It’s like they didn’t even understand the disease. Hypoxemia without patient distress.

  27. anon says:

    The Lancet paper has been withdrawn due to shoddy science!
    Curious that you didn’t mention it…

    1. Derek Lowe says:

      Do both of us a favor: have a look at both posts where I mention it before you rake me over the coals.

  28. Marko says:

    Has Trump been touting convalescent plasma ? Because it looks like the headline writers have a similar hydroxychloroquine-type bias :

    “No benefit of convalescent plasma in COVID-19 patients, study finds”

    The results shown don’t justify such a headline at all. For patients that didn’t already have one foot in the grave , there was clear evidence of benefit.

    1. loupgarous says:

      The headline clearly isn’t CIDRAP editorializing on that study’s results, merely reporting the conclusions the study authors reached. And the final section of CIDRAP’s summary went out of its way to report ‘potentially hopeful’ results and calls both by the study authors and outside commentators for further studies of convalescent plasma.

      1. Marko says:

        “The headline clearly isn’t CIDRAP editorializing on that study’s results, merely reporting the conclusions the study authors reached. ”

        Right. CIDRAP is doing exactly what Lancet did with the Surgisphere propaganda. A little editorializing about reality would have been beneficial in both cases.

        1. psoun says:

          CIDRAP are all over the place on this. See e.g., their early hating on masks and cheeky commentary: “Masks may confuse that message and give people a false sense of security. If masks had been the solution in Asia, shouldn’t they have stopped the pandemic before it spread elsewhere?”

  29. RA says:

    Most of the critiques of this paper seem to focus on internal validity. I think the internal validity isn’t bad and the nit-picking being done on it doesn’t mean much because it’s the external validity that’s more the issue….that is, the study population makes the reasonably valid results not particularly generalizable to most of the population.

    As noted in the post, those enrolled were relatively healthy. In fact, only 2 patients in the whole study got hospitalized! Median age about 40. If you look at Table S5 in the supplementary material, we see that the proportion of black and Hispanic patients in the study is quite low compared to the US population (Hmmmm….why is this buried in the supplemental material and not in Table 1?) And while educational attainment/socioeconomic status is not reported, the recruitment of health care workers and those from social media would suggest that the population is more educated and of higher socioeconomic status than the population as a whole.

    Given that COVID is more likely to progress to serious disease in those who are older, have more chronic health conditions, minority populations, low socioeconomic status populations there are limits to the generalizability of this study.

    If you are “pro-HCQ” you can note that this lower risk population is less likely to show efficacy than a study in higher-risk groups. If you are “anti-HCQ” you can note that this population is also probably less likely to have cardiac side effects from HCQ than higher-risk groups.

    I think we need studies of post-exposure prophylaxis (or immune modulation, for that matter…we should look more at famotidine perhaps as this blog has shown!) in higher-risk populations where such intervention is likely to make a difference. And the primary outcomes should be hospitalization/death or serious extended debilitating disease, not just symptoms or pcr positivity in and of themselves. And if you are going to study something like HCQ in a higher risk group for post-exposure prophylaxis, than you probably should monitor the ECG.
    It seems like nursing homes would be the right places to do studies like this…ample post-exposure opportunities, high-risk population, capacities for ECG monitoring, etc.

    1. loupgarous says:

      Have to agree with you re: external validity of Boulware’s study. There doesn’t seem to be any there there.

  30. loupgarous says:

    The study’s the last nail in the coffin for NEJM‘s scientific credibility. I understand under the present economic circumstances paucity of any scientific work to speak of, but the endpoint not being entirely based on measurement of viral load and depending largely on patients reporting symptoms in a disease where onset of infection is asymptomatic and where in general such symptoms don’t reliably herald infection… to borrow the kids’ slang… WTF?

    My experience with reported symptoms of COVID not being reliable markers of infection included at one point a fever of 100.4° F for two days, gastric reflux (while I was taking Protonix to prevent recurrence of a stomach ulcer) which refluxed down my trachea and became a scary and painful bronchitis (which resolved on its own, thankfully) accompanied by a fever… all of which symptoms were eventually followed with a negative RT-PCR I had to have to be admitted to the hospital for a liver biopsy (itself unrelated to COVID at all, but to metastatic spread of existing cancer).

    By the standards of the Boulware paper I’d have been recorded as a positive case of COVID based on symptoms indicative of infection. Forgive me if I regard this as one more step in the printed version of NEJM‘s steady progression towards birdcage liner – not because the Boulware study failed to report a clear positive or negative outcome (that happens), but because of the study’s standards for assessing actual infection with the causative agent of COVID-19.

    First the SurgiSphere-processed study took many of us in because of its sheer size (or reported size – what may eventually turn out to be a Federal jury is still out on the actual existence of SurgiSphere’s data), and now this. Here the provenance of the data’s solid (thank God), just clearly unconvincing. Infection with SARS-CoV2 is often asymptomatic, yet symptoms were used as a marker of infection (an endpoint of the study). No matter how many people this virus has killed, it’s critically injured The Lancet‘s reputation and NEJM‘s in under a week.

  31. mz_chem says:

    From the Discussion section: “ Although PCR or serologic testing for asymptomatic infection would have added to the scientific strength of this trial, this was not possible, and we cannot assess an effect on mild or asymptomatic infections. Although a marginal possible benefit from prophylaxis in a more at-risk group cannot be ruled out, the potential risks that are associated with hydroxychloroquine may also be increased in more at-risk populations, and this may essentially negate any benefits that were not shown in this large trial involving younger, healthier participants.”

    I credit the authors for trying to discover what they could. They acknowledge the limits of their trial (in the quoted portion and elsewhere) and note that it’s not definitive and that further trials are warranted.
    There is plenty to learn from even an imperfect trial and given the urgency every researcher felt in March (and now), a well executed trial that provides some data is better than zero information.

    Right now the feedback from clinicians is that this is a mercurial disease that throws new curves on a daily/weekly basis. They are clamoring for good studies to help them at least learn some of the boundary conditions for developing treatments for their patients, staff and the community.
    Until then, even a result like this one tells them how to design the next study and adds some additional data when deciding on protocols and what to say to families and those close to positive COVID patients when they come asking questions.
    I would finally note that the testing situation still has not been alleviated and possibly even more complicated than before. That’s an entirely other issue.

  32. Anonymous Cowgirl says:

    OT, but would cromolyn sodium (Nasacrom) be a possibility? I used to take this for allergies prior to the inhaled steroid boom and was told it blocks mast cells. Not sure if this is made anymore.

    1. The End of Ze World says:

      I have seen a few suggestion to use cromolyn sodium (here for example : I had a quick look at the registered clinical trials with the drug and nothing appears for CoViD yet. You might want to call your patent attorney as soon as possible.

  33. NICK says:

    For all those who put all their chips on scientific studies and their gold standard.

    “Richard Smith: Time for science to be about truth rather than careers”

  34. Bob says:

    For anyone interested:

    I ran an analysis on Boulware’s study and compared DAY 1/DAY 2 HCQ vs PLACEBO DAY 1/2 and here is what I got.

    HCQ 17/177 (9.6%)
    Placebo 26/169 (15.4%)
    Difference between means -5.8% (95% CL -12.8, 0.01)

    Odds Ratio = 0.62
    p-value = 0.1039 which is 90% significance level.

    I also ran Day 1 HCQ versus all placebo since it does not matter what day you got placebo.

    HCQ 5/77 (6.5%)
    Placebo 58/407 (14.3%)
    Difference between means -7.8% (95% CL -16, 0.0)
    Odds ratio 0.45
    p-value 0.0629 which is 94% significance level.

    1. Bob says:

      I also performed a regression analysis on the HCQ data versus Days After Exposure. R-squared of 0.84 and an intercept of about 5% meaning if they took HCQ the day they were exposed you might potentially see a 60% reduction in cases. If you took it before exposure ….

  35. EugeneL says:

    This is defintely a better study than the Lancet one. But what I think they could have done better is to go and retrospactively test those they considered positive with antibody tests, as such tests are now readily available.
    In April and March there was a lot more cold or flu than actual covid, so they could have some false positives. It’s easy to rule (some of) them out even now.

  36. theasdgamer says:

    You can find significance in the paper, but you have to do your own analysis. If you bring up the appendix and go to Figure S-1, go to the Days Exposure. Group days 1-3 together and you get 29. The number is big enough to show significance. All of a sudden HC has significant prophylactic effect.

  37. Popper says:

    Here an independent experience of cure of the combination Azi plus Hydro …until the scientific comunity do not search some benefit on the viral shedding giving the two drugs associtation at the first fases of the covid syndrome you will test another phenomena not the cure developed by Margliese doctors…this is simply to understand

  38. Poper says:

    Here an independent experience of cure of the combination Azi plus Hydro …until the scientific comunity do not search some benefit on the viral shedding giving the two drugs associtation at the first fases of the covid syndrome you will test another phenomena not the cure developed by Margliese doctors…this is simply to understand

  39. James Kringlee says:

    The University of Minnesota study you referenced on back2facts was
    poorly done and reported on worse. The U of M reported “…no
    benefit…” The P value of the study is reported as .35 meaning, it
    has been reported, the results have a 35% chance of being random.
    This does mean the study has little value not that there was a finding
    of “…no benefit…”. The study reports “the incidence of new
    illness … did not differ significantly between those receiving
    hydroxychloroquine (11.8%) and those receiving placebo (14.3%). The
    truth is that the study, such as it is, showed a 17.5% reduction in
    progression to COVID-19 type symptoms in the group taking HCQ, (14.3
    subtract 11.8 = 2.5 divided by 14.3=17.5%) with a 65% chance the the
    17.5% reduction is not random. Their report of ” no further
    benefit…zinc” truthfully told would be taking 11 or 15 milligrams of
    zinc oxide in our study did not show any significance whereas taking
    220 milligrams of zinc sulfate with HCQ is being studied. The only
    worthwhile finding was that there were no reported heart problems
    despite taking a higher dose of HCQ than in other treatments or
    studies ( 1400 mg total day one then 600mg per day for 4 days). I
    spoke to the U of M contact person for this study Kelly Glynn but I
    never got a call back on my request for clarification from the study

    The study reports “we recruited 821” and “we enrolled 821″ Yet it
    states in the Interim Analyses section of the report ” …the sample
    size was reduced to 956 participants who could be evaluated with 90%
    INFECTIONS IN THE CONTROL GROUP”. . (my capitalization) Then they
    decided the study “was deemed futile to continue” and they pulled the
    plug on their study and misreported it, so I think.

    To:   David R. Boulware:  since your study used folate/folic acid as your placebo, You may be interested in   “The role of
    folic acid in the management of respiratory disease caused by
    CORVID-19.”  It is published on .  The following is an
    excerpt from that paper.     “The results indicated that the
    interactions between folinic acid and furin were high, but
    substantially lower than that of folic acid. In this way, folic acid
    could block the access of COVID-19 spikes to furin and prevent the
    cell entry and consequently turn-over of the virus. In summary, our
    results suggest that folic acid could be used to inhibit the furin
    enzyme. The association of folic acid with furin would affect the
    structure of the protein and consequently interfere with its
    proteolytic capability. Thus, folic acid, as a safe drug, could be
    useful in the prevention or management of COVID-19-associated
    respiratory disease in the early stages of the disease.”

    Could it be that the placebo you used was not, in truth, a placebo?.

  40. Popper says:

    new data on Hydro the editorial presentation of the study

  41. An Old Chemist says:

    This just in on 07-02-2020:

    Study finds hydroxychloroquine helped coronavirus patients survive better

    (By Maggie Fox, Andrea Kane, and Elizabeth Cohen, CNN 1 hr ago, 9:00 PM)

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.

This site uses Akismet to reduce spam. Learn how your comment data is processed.