The short version: Despite many pronouncements by people in the public light, there is no evidence to support that hydroxychloroquine is viable as a treatment or prevention against COVID-19. Back-of-the-envelope calculations would suggest that the dose of hydroxychloroquine needed to accomplish a therapeutic dose would be acutely toxic and likely fatal to the person in question. Hydroxychloroquine is an important drug for many patients, and diverting it from them for COVID-19 is unethical, especially given evidence that it does not work. The drug itself also has significant potential for adverse effects and should not be used outside of well-defined clinical indications.

I had been tinkering with a post about HCQ for a little while but then once I had felt I had gathered a level of evidence I was comfortable enough with to make a compelling case, the issue seemed to die down, and then… well… see your immediate left.

At this point, it is more than abundantly clear that HCQ is not viable as therapy or prevention in COVID-19. There are myriad reasons for this, so I just want to hit upon the most salient ones.

Lately there has been a lot of renewed fascination with one particular study that appeared in IJID in early July (2020).
I am sure that this seems very compelling to a lot of people. After all, there are stark disparities that show that the patients who received HCQ did much better than those who did not, ergo it’s gotta work, right? I’m afraid this is not so. Why?

• This study is retrospective in nature. It cannot be used to claim a causal link. At best, it can point to an association, but as the study itself points out, it needs confirmation from prospective studies (more on that in a bit).

• There is no randomization here. In other words, we have no good way of ensuring whether or not the changes we are seeing are in fact due to the use of HCQ. In general, you can’t randomize a retrospective study.

• There is no double blinding here. For a study like this, double blinding is especially important because physicians treating these patients are subject to unconscious biases that may skew who gets HCQ in a systematic way, or interpret findings in a particular way, or behave differently in the care of these patients e.g. believing HCQ works so paying extra attention to those patients.

• Patients in this study were also much more likely to have received steroids if they got HCQ, which emerging evidence is showing may have therapeutic value in later COVID-19.

• The patients who got HCQ were a priori at lower risk of death as they were not contraindicated for the drug.

• I think they may have overcorrected for some factors in their statistics because they found that being white puts patients at higher risk of death. This is at odds with virtually all the data we have on COVID-19 and the entire field of social determinants of health. It’s also possible that this represents an outlier, in which case extrapolating this data to the norm warrants special caution.

In short, this study was interesting but as the authors explicitly state, it cannot be used to guide clinical practice in isolation. I don’t want to make the claim that this was a bad study. There was a time in which we genuinely did not know whether HCQ had value in the management of COVID-19 and it was important that we gather all the data and examine it critically, regardless of what it showed.

That brings us to the more robust data:

This is a randomized controlled trial and prospective, but it is also not double blind. The results align better with what we expect based on past experience with HCQ: there were increased incidences of heart arrhythmias in patients receiving it. But there was also no significant effect (which is to say, the most probable explanation of between group differences is random chance). Additionally, it is examining mild-to-moderate COVID-19, which, while certainly serious at the level of public health and medicine, is not the major clinical entity that concerns us. After all, the major goal of therapy above all else right now is to reduce mortality rate.
Another claim about HCQ raised is that it has the ability to prevent COVID-19 if taken prophylactically.

This has been examined with the ideal type of study under the circumstances. This is a prospective, double-blind, placebo controlled trial: no effect.

As I am about to explain, this really should not come as a surprise.

A word on antivirals: it is incredibly hard to make an effective one. For one thing, most viral infections are acute in nature (the ones for chronic infections tend to be quite a bit better) and by the time the patient has symptoms the virus has already proliferated fairly extensively in most cases and so the value of the antiviral in the first place is questionable because the immune system is now angry and going to do everything possible to clear it (this is why Tamiflu is basically pointless if not given within 2 days of symptom onset). However, it’s also why Tamiflu works so much better as post-exposure prophylaxis rather than therapy (though whether this is true of a given antiviral depends on the mechanism of action). Taken together, the sobering reality is that an antiviral drug basically regardless of the mechanism, is probably not going to be very useful in reducing mortality in very sick patients. With remdesivir, which we know at this point has *some* efficacy as an antiviral, its principal benefit is it shortens the duration of illness. At this point in time, there isn’t good evidence for a reduction in mortality, though it’s possible that as new data emerge, that may change. Remdesivir’s major benefit as I can see would have to do with the capacity of the healthcare system to handle cases: if people get ill and get over it more quickly, you have a more rapid turnaround of beds and therefore more capacity to deal with patients. The extent to which this may be important is unclear. Tamiflu is similar in this regard: it likely does prevent some cases of influenza pneumonia based on its mechanism of action but empirical data for that are lacking and thus far the only thing we can say with a high degree of confidence is it shortens the duration of illness. That’s why if you poll a bunch of physicians who deal with influenza patients regularly, you will see a great deal of controversy in their opinions. But back to the point: HCQ.

So, if HCQ works so well in vitro, why doesn’t it seem to work all that well in vivo? Firstly, let’s talk about what HCQ does. Principally, HCQ is an immunomodulating agent used in rheumatic diseases, especially lupus because it does a few things: firstly, it directly inhibits a protein called TLR9. TLR9 is a sensor of DNA, specifically CpG DNA, and it can set off potent inflammation. This is particularly important in lupus, where host DNA can become the target of aberrant immune responses. The other major thing it does is it interferes with antigen presentation. Without overcomplicating it, antigen presentation has intimate requirements of an organelle called the lysosome. The lysosome is sort of like the cell’s incinerator. It’s a very acidic compartment filled with enzymes that can digest almost anything. In fact we don’t really get for sure why lysosomes don’t digest themselves. The thing is: for these enzymes to work, they need that acidic pH. HCQ is a weak base and it’s extremely lipophilic. That means it has a very easy time getting into cells, and it gets trapped in the acidic compartments. However, when acid and base encounter each other, neutralization occurs, and that creates a problem for those enzymes because they don’t function effectively at neutral pH. Now, where does the virology come in? Viruses can enter cells through a few ways, but one of the principal ones is endocytosis. They interact with a receptor on the cell’s surface, the cell, surrounds them with their membrane, and then tries to treat them like food: it fuses the membrane of the compartment containing virus and a lysosome to break them down into raw material that the cell can use. The issue is, viruses are very clever and have evolved to exploit this as a mechanism to enter cells e.g. the enzymes cleave fragments off viral proteins that activate those proteins and enable release of the virus’s genetic information into the cell’s cytoplasm (or in some cases, the acid from the compartment itself activates the viral proteins to facilitate invasion). So, if you can stop that process, for some viral infections, the virus won’t be able to enter cells.

Here’s the problem though: the concentration of HCQ needed to do this is in the micromolar range. Accomplishing this locally in a petri dish of cells is not a problem. Accomplishing this in a human is another story. Let’s suppose for a moment I administer HCQ intravenously, so 100% of it ends up inside the body. There are about 6 liters of blood in your typical adult, but the volume of distribution for HCQ is about 6000 L because it binds many proteins and enters cells in the plasma before it can reach its target site. To attain a clinically significant finding, the general rule of thumb in pharmacology is you need 10–100 times the IC50 or EC50 (depending on what the function of the drug is). I’m going to use very round numbers here and I’ll call EC50 2 micromolar based on the estimates provided in the above-linked in vitro study. That means I need ~12000 micromoles of hydroxychloroquine in a person. This is equivalent to ~4.03 g of HCQ to attain something resembling the EC50 (for an intravenous dose). For significant antiviral effect (per our rule of thumb), that’s about 40 to 400 grams. The dose we would use for rheumatoid arthritis management is 400–600 mg per day, roughly one-thousandth to one-hundredth of this.

There’s actually another reason we know that this drug cannot possibly work in humans in either a therapeutic or prophylactic capacity. You can read the explanation here, which is superb:

In short, there is yet another pitfall here as far as the research goes: the choice of cell type in your in vitro studies. The studies which demonstrated an effect with HCQ/CQ are studies performed on Vero cells which have been genetically modified to express ACE2, the receptor for SARS-CoV-2. When there is ACE2 present on the surface of the cell, but NOT TMPRSS2, an enzyme that cleaves the spike (S) protein of SARS-CoV-2 into is active form, the virus enters through the endocytic pathway described above with all those lysosomes and endosomes. In this pathway, we would expect it to be vulnerable to the action of hydroxychloroquine (assuming a reasonable dose were possible). Here’s the issue though: we are not made of monkey kidney cells. When you attempt to reproduce this effect in respiratory epithelial cells, which are the major target of SARS-CoV-2, this effect is abolished. That’s because respiratory epithelial cells express TMPRSS2, which cleaves the spike protein into an active form that can induce fusion of viral membranes with that of the cell in a manner completely independent of the endocytic pathway. When you compare HCQ and CQ to an inhibitor of the TMPRSS2 protein, camostat, you see in vitro a very strong inhibitory effect of camostat, and virually none from HCQ or CQ until you hit absurd doses.

HCQ has significant risks to consider as well and while the adverse effects are not particularly common, giving it to enough people, especially those who do not need it, will cause harm.

I really wish there were a simple answer like this, but there isn’t. We have a history of trying HCQ as an antiviral it just doesn’t work.

The amazing people at science based medicine also have an excellent discussion of the subject of HCQ.

In vitro and in vivo efficacy are very different things and just because you have in vitro efficacy does not mean you have in vivo efficacy. In fact, it is generally going to be the case that you don’t. Between the dosage you would require in theory based on in vitro studies, the clear inability of HCQ to even demonstrate any antiviral effect in appropriately selected cells, and the increasing bodies of negative data, it is abundantly clear that HCQ does not have viable therapeutic or prophylactic utility in the management of our current pandemic. I know that this issue has become very politicized, but I implore everyone: look at the data and see reason. We need to devote our resources to more fruitful avenues and we need to have the patience to do good science.