Several people have contacted me over the last few weeks asking for my opinions on hydroxychloroquine. I’ve invariably answered that I don’t know, since I haven’t looked at the data myself. I felt that it was time to rectify that situation.
As many people know, hydroxychloroquine is an old drug that has existed since the 1950s and that is primarily used to treat malaria and rheumatological disorders. It is a close chemical relative of quinine, which has been used to treat malaria for centuries. It is also a so called “dirty drug”, by which is meant that it has multiple effects in the body that aren’t fully understood. This is in contrast to most modern drugs, which are generally required to have a very specific and well understood effect in order to be approved for use in patients.
The first study on hydroxychloroquine for treating covid came out in March in The International Journal of Anti-microbial Agents. It was an open-label, non-randomized clinical trial involving 42 patients who had tested positive to covid-19 with a nasal swab. Open-label means that both the patients and the treating doctors knew who was getting the drug and who was in the control group. Non-randomized in this instance means that the patients were allowed to decide for themselves which arm they would be included in – the ones that wanted to try hydroxychloroquine got it, while the ones that didn’t want to try it ended up in the control group. There weren’t quite enough people to fill out the control group through this method, so it was further filled out by patients in another hospital where hydroxychloroquine wasn’t being used. The study was funded by the French government.
26 patients received hydroxychloroquine, while 16 patients served as a control group. Six of the patients in the hydroxychloroquine group dropped out of the study for various reasons and weren’t included in the analysis. This is actually a big problem, because it means that people who were potentially dropping out because of side effects of the drug weren’t included in the analysis, which can make the drug look safer and better than it is. No-one in the control group dropped out of the study.
There was quite a bit of variation in how sick the different patients were; among the 36 patients that were included in the analysis, six were completely asymptomatic, while 22 had symptoms of an upper respiratory tract infection (for example a sore throat or a runny nose), and eight had symptoms of a lower respiratory tract infection (for example trouble breathing or a high temperature). The average age of the patients was 45. The hydroxychloroquine group was a bit older than the control group (51 years vs 37 years), and a bit sicker (30% had a lower respiratory tract infection vs 13% in the control group).
The main thing the researchers were looking at was whether or not there was a change in positivity to covid-19 on a nasal swab. This is another problem with the study, because the nasal swabs are not particularly reliable (which we’re going to go in to more detail on as we discuss the next study). At the start of the study, everyone had a positive nasal swab, since that was a criterion for being included. At day 3, 50% of patients in the hydroxychloroquine group had a negative nasal swab, compared with 6% in the control group. At day 6, 70% of patients in the hydroxychloroquine group had a negative nasal swab, compared with 12.5% in the control group.
What conclusions can we draw from this? The first thing to say is that this was a really low quality study. The fact that there was no blinding and no randomization meant that there was a huge scope for confounding effects. Were the doctors treating the patients with hydroxychloroquine differently from the controls in some way? Were they as thorough when taking nasal swabs from people in the hydroxychloroquine group as from people in the control group? (if you don’t stick the swab far enough back in the nose, and jiggle it around enough, you’re more likely to get a negative result). Were the patients who willingly said yes to hydroxychloroquine different from the controls in some significant way? Or did they act differently from the controls in some other way that affected their disease course? Without blinding and randomization, it’s impossible to know and impossible to get rid of these types of confounding effects.
The second big problem with the study is that it was tiny – only 36 patients were included in total, so the scope for random variation to mess up the results was huge.
And the third big problem is, as mentioned, that six patients treated with hydroxychloroquine were removed from the analysis after inclusion in the study. This is known in the scientific jargon as doing a “per-protocol” analysis, and it is heavily frowned upon because of its ability to distort results.
Basically, this was a really crappy study and we can’t draw any conclusions either for or against hydroxychloroquine from it. I’m surprised it received as much attention as it did when the results first arrived – the only explanation I can see is that at the early stage of the pandemic everyone was so desperate for any guidance that they were willing to look at even the lowest quality evidence. All we can say from this study is that there seems to be some merit to hydroxychloroquine, and that it is worth doing further studies.
Another study was published in the British Journal of Medicine in May, looking at the use of hydroxychloroquine in patients with mostly mild to moderate disease. Like the previous study, it was open label, but unlike it, the patients were randomized to treatment or placebo, rather than getting to choose for themselves. The study was funded by the Chinese government. 150 patients were included, 75 of whom received hydroxychloroquine in addition to standard care, and 75 of whom just received standard care. The average age of the patients was 46 years. 148 of these patients had mild to moderate disease, and only two had severe disease. Basically, if patients didn’t need oxygen supplementation then they were classified as having mild to moderate disease, so this wasn’t a particularly sick group of patients. Only two patients out of 150 required oxygen, and not a single patient died.
On average, hydroxychloroquine was given 17 days after the start of symptoms. This is an important detail to consider, since people generally develop symptoms five days after being infected, which means that the average person was getting hydroxychloroquine around 22 days after being infected! To me that sounds really late. By that point, I would think that most people would have cleared the infection already. Since the main hypothesis behind hydroxychloroquine’s ability to fight covid is that it interferes with the virus’s ability to replicate, that would seem to doom the study from the start.
In order to be eligible, patients had to be over 18 years old, healthy, and have a nasal swab (or alternately a sample from the lower airways) positive for covid-19. Diagnosing active infection with a nasal swab is problematic, since we now know that people can continue to be PCR positive for covid for two or sometimes even three months after infection, which means that a positive nasal swab can very well be a sign of an old infection that has already cleared up. In that case, treating with hydroxychloroquine obviously isn’t going to be helpful, which could make the results seem weaker than they are. (PCR stands for Polymerase Chain Reaction. It is a method in which a segment of DNA or RNA is duplicated many times until there is a large number of copies of that segment. If the segment you are looking for is actually in the sample, even if only in very small amounts, it will eventually become noticeable after 30 or 40 cycles of duplication).
As in the previous study, the researchers were trying to determine if hydroxychloroquine was having an effect on the infection by doing repeat nasal swabs, and checking if they were still positive. Knowing what we know about how long people can continue to be PCR positive after having cleared the infection, we can almost say for certain before we even look at the results that they’re probably not going to tell us very much.
Anyway, let’s see what the results showed. On average, it took the patients receiving hydroxychloroquine eight days to have a nasal swab negative for covid, while it took the control group an average of seven days to have a negative nasal swab. At the 28-day mark after start of treatment, 85% of patients in the hydroxychloroquine group had a negative nasal swab, compared with 81% in the control group. Furthermore, at 28 days, 60% in the hydroxychloroquine group were free of symptoms, compared with 67% in the control group. The median time to symptom resolution was 19 days in the hydroxychloroquine group and 21 days in the control group. None of these minor differences was statistically significant.
In the hydroxychloroquine group 30% had some kind of adverse event, compared with 9% in the control group. This is as would be expected. Hydroxychloroquine has known side effects, of which stomach pain, vomiting, and diarrhoea are the most common, so it would be strange if there weren’t more adverse events in the treatment group than in the control group.
So, where are we so far? Basically we have two studies that, because of the way they were designed, don’t tell us anything useful. The first was too low quality to give any real insight, and the second gave hydroxychloroquine far too late in the course of the disease to have any meaningful impact.
Next we have a study that was published in The Annals of Internal Medicine in July. This was a double-blind randomized controlled study (aren’t those just the most beautiful five words in the English language?). 491 patients with mild to moderate covid-19 (i.e. not requiring hospital admission) were randomized to receive either hydroxychloroquine or placebo. I say that the patients had covid-19, but due to shortages of tests in the US at the time the study was performed, only 58% actually had a positive nasal swab. The rest were assumed to have covid-19 due to their symptoms and proximity to someone with confirmed covid-19. This is of course a weakness with the study, since we can’t actually be completely certain that 42% of the participants had covid.
When it comes to funding, the authors state simply that the study was funded by “private donors”, which makes it impossible to assess if there was any specific pro- or anti-hydroxychloroquine agenda behind the study. It’s always good to know what the authors of a study wanted the results to be, since that makes it easier to see if they are presenting their results honestly or trying to massage them in some way.
Unlike the previous two studies, the investigators weren’t looking at the results of repeat nasal swabs to determine whether the drug was having an effect, but rather were looking at what symptoms the patients had. This is a good thing because it’s what matters most – patients don’t care about whether a test is positive or negative as much as they care about how they feel. Symptom severity was assessed using a ten point scale, where zero indicated no symptoms and ten indicated the severest symptoms imaginable.
The median age of the patients was 40 years. In order to be eligible for the study, patients had to have had onset of symptoms less than four days previously. 56% of the patients were enrolled in the study within one day of symptom onset. This is another good thing with the study. If the hydroxychloroquine is going to have any chance of having an effect, it most likely needs to be given early.
At enrollment in the trial, 98% of participants had symptoms. On the fifth day, 54% in the hydroxychloroquine group had symptoms, compared with 56% in the placebo group. At the two week mark, 24% in the hydroxychloroquine group had symptoms, compared with 30% in the placebo group. Overall, symptoms decreased by 2.60 points on the ten point scale in the hydroxychloroquine group, and by 2.33 points in the placebo group over the course of the two weeks. The small difference between the groups was not remotely close to being statistically significant. The results did not change when those with PCR-confirmed covid were analyzed separately from those without confirmed covid.
As in the previous study, one difference that was statistically significant was the rate of adverse events. 43% in the hydroxychloroquine group had an adverse event, compared with 22% in the placebo group.
What conclusions can we draw? First of all, this study is of much higher quality than the previous too, and the hydroxychloroquine was given in a timely manner after infection, and a patient oriented outcome was studied instead of just a lab value. Therefore, its results are much more scientifically valid than the previous two studies’ results. And those results say that hydroxychloroquine doesn’t work as a treatment for covid, at least not in people with mild to moderate disease. If it does have an effect, that effect is at best marginal, since this study, involving over 400 people, wasn’t able to see any meaningful difference, and needs to be weighed against the increased frequency of adverse events.
My main complaint with this trial is that the patients were relatively young, and not particularly sick. A trial in which the patients were over 70 and living in care homes would have been more interesting, since that is the group most at risk of dying from covid, and for which it is most urgent to find an effective treatment.
A similar study was published in Clinical Infectious Diseases, also in July. It was an open-label randomized controlled trial that included 293 patients with mild covid-19 and less than 5 days of symptoms. The funding for the study came from a crowdfunding campaign, and the mean age of the participants was 42 years. 136 patients received hydroxychloroquine while the remainder received standard care.
One interesting thing about this study is that it actually measured viral load instead of just checking whether patients were covid-positive or negative. There was no difference in viral load at either 3 days or 7 days after the start of treatment. The average time to recovery was 10 days in the hydroxychloroquine group and 12 days in the standard care group, a minimal difference that wasn’t anywhere close to statistically significant. 72% of patients in the hydroxychloroquine group had at least one adverse event, compared with 9% in the standard care group.
So, again, hydroxychloroquine doesn’t seem to work as a treatment, at least not in relatively young patients with relatively mild disease. Could it at least work for prophylaxis, i.e. to prevent infection?
Luckily there is a trial that sought to answer exactly that question. It was published in the New England Journal of Medicine at the beginning of August, and it was a double-blind randomized controlled trial.
821 individuals were recruited in to the study. In order to be eligible they had to have spent at least ten minutes within six feet of a person known to be infected with covid-19 during the previous four days, and not started showing any symptoms at the point when they started treatment with hydroxychloroquine or placebo. Again the average age of the participants was around 40.
Among the people who received hydroxychloroquine, 11,8% went on to develop symptomatic infection. Among the people who received placebo, 14,3% went on to develop symptomatic infection. This is an absolute difference of 2,5%, but it was nowhere near statistically significant.
Conclusion: Hydroxychloroquine does not appear to have any useful role in relation to covid-19, based on the best evidence available in the peer-reviewed literature at this point in time (August 2020). That conclusion may change as more evidence becomes available (there are multiple pre-prints awaiting peer-review and also a number of trials ongoing), but at the moment, it doesn’t make sense to treat patients with hydroxychloroquine outside of randomized controlled trials.
One thing I think is important to add is that these studies were all looking at relatively young people, most of whom had mild to moderate disease. It is still possible that hydroxychloroquine has a role in treating older, sicker patients, and it is my hope that additional trials will soon be published showing what the effect of hydroxychloroquine is in this group. However, one thing that has also become clear from the above trials is that hydroxychloroquine has a high frequency of side effects. For an old, sick person who is teetering between life and death, taking a drug that causes them to develop diarrhoea could be the thing that pushes them over the edge and kills them. That is why no-one should be taking these drugs outside of clinical trials.
If you thought this article was interesting, then you may also enjoy my article about how to determine immunity to covid-19 in a population with T-cells or my article about whether you should take fever lowering drugs when you have an infection.