Dr. Didier Raoult of Marseilles and his co-workers have published another preprint on clinical results with the chloroquine/azithromycin combination that their earlier work has made famous. And I still don’t know what to think of it.
This is going to be a long post on the whole issue, so if you don’t feel like reading the whole thing, here’s the summary: these new results are still not from randomized patients and still do not have any sort of control group for comparison. The sample is larger, but it’s still not possible to judge what’s going on. And on further reading, I have doubts about Dr. Raoult’s general approach to science and doubts about Dr. Raoult himself. Despite this second publication, I am actually less hopeful than I was before. Now the details.
I. This Latest Study
The new manuscript calls this “an observational study”, but I don’t see how that’s right – this is interventional, and that’s the whole point. In it, the Marseilles team reports treatment of 80 patients. Median age was 52 years (range 20 to 86), nearly 1:1 male/female. Six of the patients were from the earlier study reported by the group, here with longer follow-up. The patients were grouped into those with upper respiratory symptom and those with lower (54% of them in that category), and comorbidities were noted (46 of the patients had at least one known risk factor) as well as days between onset of symptoms and hospital admission, and between admission and start of treatment. Interestingly, only 15% of the patients had a fever, and four of them were noted as asymptomatic carriers, which makes me wonder how they got them into the study in the first place. Patients were give a CT scan shortly after admission to look for pneumonia – the body of the article says that patients “systematically underwent” this test, but the data tables show that 20% of the subjects were not evaluated. Of those who were, 54% had radiologic pneumonia signs. Nasal swabs were taken daily (“with some exceptions”) and analyzed for viral RNA via RT-PCR. Cultures, meanwhile “were attempted in a random selection of patients”.
Patients received hydroxychloroquine (200mg t.i.d for ten days) and azithromycin (500mg on the first day, 200mg q.d. for the next four days). The patients with pneumonia and bad overall “national early warning scores” (22%) got additional antibiotic (ceftriaxone, dose and schedule not specified). Patients had an ECG before treatment and two days after starting treatment, looking for any signs of QT interval prolongation. Treatment was discontinued (or not started at all) depending on ECG results, and any other drugs that are associated with QT prolongation were discontinued. The paper says that symptomatic treatments, including oxygen, were added as needed.
The end points of this treatment were messy, probably unavoidably so:
Criteria for discharge changed over the course of the study. Initially, patients with two successive negative nasopharyngeal samples resulting from PCR assay (CT value ≥35) were discharged. From 18 March, patients with a single nasopharyngeal sample with a PCR CT value ≥34 were discharged to their homes or transferred to other units for continuing treatment, Ultimately, because of a crucial need to admit new, untreated inpatients, inpatients already receiving treatment with a PCR CT value <34, with good clinical outcome and good adherence to treatment were also discharged.
OK, that sets the stage. We’ll come back to the discharge numbers in a bit. The study’s results include all the patients who got the HCQ/AZ treatment for at least three days and were followed up for at least six. A lot of the data are included in the following chart
So you can see that by day 3, about five patients had left the study, and by day 5 there were 61 patients still having data collected, and so on. The tan-colored bars are the PCR values from the nasal swabs; negative results are defined as anything that hit 35 or above. So one thing you can see is you start off with all 80 patients at 34 or under, which is what you’d expect, since it’s one reason that they were admitted into the study. On day 2, 10 patients now showed a negative nasal swab test (70 patients still positive). Now, according to the paper’s Table 2, 49 patients started therapy on Day 0, and 26 on Day 1. So this apparently means that 10 out of the 49 got to a completely negative nasal swab reading overnight? That seems hard to believe, unless they were already close to the cutoff, but here’s a big problem with the preprint: we do not have individual patient data. The previous smaller study from the Marseilles group included a table of such data (which had some inconsistencies, mind you) but it is not present here, and I can only hope that it shows up in the final published version. Update: Lu Chen has followed the numbers all the way down and finds more inconsistencies.
To see even more why that’s needed, let’s go out to Day 2. By this point, they’re still testing all 80 patients (the black bar) and now, what, about 64 of them are still positive (tan bar)? So of the 39 patients that started on Day 0 and the 26 that started on Day 1, only 6 more of them improved enough on the nasal swab to be called negative for viral RNA. In other words, 20% of the patients treated went clean during the first 24 hours, but less than 10% did over the second 24. The “number of patients” bars look pretty reasonable, but that’s before you look at who started treatment and on which day. And wouldn’t it be useful to see the progress of the individual patients day by day in that PCR test? How the most heavily viral-loaded ones did compared to the lighter ones, how the asymptomatic carriers did compared to the rest? We can’t. All we have are the aggregate numbers.
And we’re missing a very important aggregate number indeed: a control group. How would a comparable group of patients have performed in these RNA tests for contagiousness under another standard of care? Even with or without azithromycin, if you can’t stand the thought of not giving them hydroxychloroquine? We don’t know. Without matched controls, and without being able to look at individual patient data, we just don’t know how good this treatment was or frankly if it was any good at all. We may be seeing a notable effect size in what is still a small trial, or we may be seeing something that’s not that remarkable or the result of a poorly controlled protocol. We don’t know. I understand the need for speed, and I’m glad that the Marseilles group is conducting studies and releasing them as preprints. But this work does not help us anywhere as much as it should.
Now, many of the patients in this study were, in fact discharged, but not all (65 of the 80). Three patients were transferred to ICU during the study – two improved and came back to the main ward, and one was still in ICU at time of publication. One elderly patient (86) died before going to the ICU – that one is the only age associated with a specific outcome in the whole paper, consistent with the lack of individual data. How do these figures for disease progression compare with other treatments for similar patients? Who knows? This is an interesting question when you consider this 2016 letter from Dr. Raoult in the journal Clinical Infectious Disease, pointing out the need for strict negative controls when evaluating viral etiology. However, he is at least true to his view from this interview, where he mentions that he does not believe that randomized trials are useful in infectious disease work.
II. Previous Work
That takes us to Dr. Raoul’s other published work. For extended comment on this I refer the reader to this post by Leonid Schneider at For Better Science. To summarize, there are a number of papers published from his lab over the years that have some of the better-known publication sins: duplication of photomicrographs, photoshopped blots. One of these in 2006 was egregious enough that Raoult and several of his co-authors were banned from publishing in any ASM (American Society for Microbiology) journals for a year. He was angry enough about this that he has almost never published in an ASM journal since the incident.
I am a believer in the maxim that you should never ascribe to malice what can be explained by incompetence. When you see these sorts of things in a publication, it can be outright fabrication, or corner-cutting (not permissible either, of course), or sheer disorganization and sloppiness (which also not should be the case). Raoult publishes a lot of papers (hundreds of them), and I suppose one shouldn’t be surprised that there’s some junk in there. I don’t think he rises to the level of some serial fraudster. But neither does this stuff build confidence.
It’s also interesting to take a look at his earlier reactions to this very epidemic. In this YouTube video from January 21st, he takes a rather dismissive tone (translation from a transcript here, in a long article (in French) at Les Crises that serves as an excellent source of background on Dr. Raoult in general):
Q: Prof. Raoult, a coronavirus epidemic is making the news in China. Do we have something to fear?A: You know, it’s a crazy world. What’s going on, the fact that people have died of coronavirus in China, you know, I don’t feel very concerned. It’s true that the world has gone completely mad: if something happens where 3 Chinese people die and it makes the world news, WHO gets involved, it goes on the radio, on television. If there is a bus crash in Peru we say “road accidents are killing more and more people”. All this is crazy. That is, there is no longer any clarity .Whenever there is a disease in the world we wonder if we are going to have it happen here in France. It just becomes totally delusional. . .I don’t know, people don’t have anything to do, so they go to China to find something to be afraid of. . .well, it’s just not serious.
To be sure, he has changed his mind on the subject, as have many others. But that should relieve him of any duties as a prophet people might want to assign. You say, that’s from late January, and we shouldn’t hold people to what they had to say back then? Well, here are some of his statements in an interview with an Italian magazine on February 24th:
You know, there are more deaths from scooter accidents in Italy than from the coronavirus. This psychosis and runaway media come from a sensitivity of the human race to the risk of extinction. Anthropologically, there is always a “Reason We Are All Going To Die” . . .It turns out that epidemics are part of the modern world, so that necessarily causes concern. . .De facto, if we look at the numbers, Monday February 24, 2020, there were only 500 new coronavirus cases in the world – these figures do not justify this massive panic. Every year there are tens of millions of deaths in the world due to viral respiratory infections; there will be a few hundred more.If you look at the new cases, the rate of new contamination is less than 1% right now; it’s very low and it suggests that the epidemic is coming to an end. . .
III. Dr. Raoult Himself
As to that personality, I don’t know what to think of Dr. Raoult himself, either. His public statements about his recent work have not lacked for confidence – over and over, he has said that he believes that he has a cure, that there is no reason for any physician not to start prescribing his combination immediately and that it would be tantamount to malpractice to do otherwise.
But he is a man of strong opinions. Update: see this profile in Science from 2016 for more of these. Unfortunately, he has also seen fit to share them in a quickly written book, Épidémies: Vrais Dangers et Fausse Alerts (Epidemics: Real Dangers and False Alarms), which comes complete with a large photograph of Dr. Roualt on the cover. It is hard not to find this unseemly in the current situation, and if there is a provision for the profits to be donated to charity, then I have missed it. You might take away from all this that a large ego is at work, and the Inspector General of Social Affairs of the French government would agree with you on that point. A 2015 report on the management and structure of the IHU-Marseilles, Dr. Raoult’s institution, concluded that it was far too centralized on his decision-making and “entirely dominated” by him in both the administrative and scientific senses.
Raoult has, in fact, been the subject of a number of accusations of harassment over the years. The For Better Science post linked to above has more details, and if you read French or want to use Google Translate you can see quite a few of them here. There’s a lot of “Who told you you could speak up” and “You’re not paid to think” browbeating, complaints of shouting matches and arbitrary dismissals, etc., which do seem to fit what one gets of the man’s personality in his interviews. Update: Leonid Schneider has heard from someone who worked in Raoult’s institute, and who vividly describes the atmosphere there. Complaints of sexual harassment also boiled over in 2017 with at least six women involved. Allegations were made that another professor at IHU threatened a foreign graduate student and others with retribution if they came forward against him and made statements like “Don’t you know how to behave with white people?” after making sexual advances. Prof Raoult was accused of trying to minimize the affair in the department and of threatening to fire people who were making accusations. It is not comfortable reading. (Update: cleared up this section and added more details).
All in all, I am pretty sure that I don’t care for Didier Raoult very much. And I don’t care for his style of research nor for his ways of expressing himself. Now, it would be a more simple world if assholes were always wrong about things, and I am not yet prepared to say that Dr. Raoult is wrong about hydroxychloroquine and azithromycin. But neither does he seem to be the sort of person who is always a reliable source, either. I do not take pleasure in this. But I am less hopeful about this work than I was when I first read about it, and I can only wonder what direction those hopes will take in the weeks to come.
Update: more from David Gorski at Science-Based Medicine. The words “nothingburger” and “useless” make an appearance in his description of the latest paper.
Update: more at Slate, who refer to Raoult as “Trumpian”, which I have to say is an adjective that I had considered as well.