There are several clinical trial updates for today – some that have reported data, and two that are apparently never going to report at all.
I refer to trials of Gilead’s remdesivir in China, one for moderate cases and one for severe ones, both of which were updated yesterday to “suspended” on ClinicalTrials.gov. That took people by surprise, and Gilead’s stock dropped in response. Both trials have the notice: “The epidemic of COVID-19 has been controlled well at present, no eligible patients can be recruited”. Now that’s interesting, and the first thing to note is that these were not trials that had been initiated by Gilead themselves. Here’s the company’s list of remdesivir trials, and these are the ones shown as initiated by Chinese health authorities, the ones coordinated by the China-Japan Friendship Hospital. For their part in these, Gilead contributed remdesivir at no cost and advised on trial design (both were randomized, double-blinded, and placebo-controlled).
But that means that these same authorities were responsible for suspending the trials, so what are we to understand from that? Many analysts have speculated that if the data were better for the drug that this wouldn’t have happened, and we can’t rule that out. But I also wonder about the explanation that no eligible patients can be recruited. Has the epidemic really receded that totally in Hubei? Or are patients just unwilling, for one reason or another (perhaps it’s the placebo control), to join these trials?
Update: I’m told that one big issue was the stringent inclusion criteria for the trials – apparently patients had to have no previous therapy with any other experimental agent to enroll, and that eliminates a *lot* of people.
According to the listings, the severe-patient trial had been enrolling patients since the first week of February and shows an Actual Primary Completion Date (the date on which the last patient received a trial intervention) of March 30, and an Actual Study Completion Date (the date on which the last patient had data collected) of April 10. So it would seem, at least from way over here, as if they had collected a fair amount of data (I have seen a figure of 237 enrollments out of a target of 300 patients) which makes you think that the speculation about lackluster effects might not be completely off the mark. Could they have unblinded right after that last data collection and decided that there was nothing to go on with? That seems like a quick decision, if so, but these are strange days. At any rate, given that remdesivir has not shown any particularly compelling data so far, I would say that people who have their hopes up for it should update to a new posterior probability, as the Bayesians say.
And while we’re on the topic of updating one’s expectations, there are more studies of hydroxychloroquine to review. This one (a multicenter effort from China) is randomized and has a control group, although it’s open-label. 75 patients received standard-of-care, while another 75 received that plus a loading dose of 1200 mg HCQ/day for the first three days, followed by 800 mg qd for two weeks (or three weeks in severe cases). Patients were 18 and over, were checked by upper respiratory swab (RT-PCR assay) for viral RNA on admission to the trial, and again at days 4, 7, 10, 14, 21 and 28. The primary endpoint was the rate of negative swab tests at day 28, and the secondary endpoints were negative tests at the earlier time points, improvement rate of clinical symptoms within the 28 days, and normalization of C-reactive protein (CRP) and lymphocyte count within that same period.
That primary endpoint was not met: the number of negative tests at the 28-day mark was statistically indistinguishable between the two groups at the 28-day mark (both in the mid-80% range). Post hoc analysis did not identify any subgroups that showed a difference, either (age, gender, obesity, pre-existing conditions, severity of disease, etc.) As for the secondary end points, (1) the rates of negative conversion at all the earlier time points were indistinguishable as well, (2) the rates of alleviation of symptoms between the two groups were likewise indistinguishable at 28 days – a post hoc analysis controlling for dosing of other anti-viral agents suggested more improvement in the HCQ group, but this did not reach statistical significance, (3) reduction in CRP and lymphocyte values seems to have been slightly more rapid in the HCQ-treated group (the two diverged a bit around day 8, although the two groups became were back to indistinguishable by day 14). In other words: no significant differences were seen with hydroxychloroquine treatment. There were no significant adverse events during the trial, but nothing that was apparently helpful, either. I have to congratulate the team that got this study working under what were very difficult conditions and their ability to coordinate so many testing centers. I’m sure that they would have wanted more exciting results, but the data are the data, and they’ve done valuable work by collecting it.
The second study is from France. It’s a retrospective one, not an intentional clinical trial, so keep that in mind. The team went back through the recent records across four French medical centers to compare 181 total coronavirus patients who were receiving oxygen as part of their therapy. 84 of these received hydroxychloroquine (600mg qd) and the other 97 did not; standard of care of the two groups was otherwise similar, as was their severity of disease. The composite primary endpoint was transfer to an ICU within 7 days or death from any cause, and a secondary endpoint was the development of acute respiratory distress syndrome (ARDS).
There were no statistical differences between the two groups in that primary endpoint, unfortunately, nor in the rate of developing ARDS. Meanwhile, 8 of the HCQ-treated patients had electrocardiogram changes that required them to be taken off the drug. The authors conclude that these results “do not support the use of HCQ in patients hospitalised for a documented SARS-CoV-2 pneumonia”.
Next up is another retrospective study from Guangzhou and Hong Kong, examining the records for 284 consecutive patients admitted in Guangzhou. 89% of them cleared viral RNA by day 21 overall, but there was no effect seen when examining any subgroup of antiviral treatment: oseltamivir (which one would not expect to work, a priori), arbidol (which to my knowledge has never shown activity under any controlled conditions, to the point that its Russian manufacturer seems to have actively avoiding such studies), the lopinavir/ritonavir combination (which has since failed at least two controlled coronavirus trials). . .and chloroquine. No effect. 43% of the patients received none of these drugs, while the others had them singly or in various combinations, but nothing differed from the supportive-treatment-only group.
Now, there are many more trials underway, and I very much look forward to their readouts. But when you look over the actual controlled data that we have so far for hydroxychloroquine, those previous links plus what we have today, the case for the drug is not encouraging at all. There is one small trial (from China) that showed some positive results, and data from China, Brazil, and France that show no benefit for either hydroxychloroquine or chloroquine itself and (in some cases) evidence of actual harm. The trials that are yet to report are going to have to start showing some strong positive effects if this story is going to have a good ending.
I get plenty of email and Twitter static to the effect that I’m favoring on-patent compounds like remdesivir over cheap generics like hydroxychloroquine, so let me reiterate the point I made earlier in this post: remdesivir isn’t looking very impressive, either. There is as yet no reason to recommend its use in the Covid-19 epidemic, either, as far as I’m concerned, and its ongoing trials are going to have to really look different from what’s come before, too. Honestly, at the moment, I don’t see any small molecule antiviral therapies that have much to recommend them based on the evidence in hand. Let’s hope that changes; there are quite a few being evaluated, and we really could use some encouraging data.
And while we’re mentioned email and Twitter static, the number of people who are telling me that I need to believe Dr. Raoult and the Marseilles group and that we need to immediately give everyone hydroxychloroquine, well. . .they’re beyond counting. But here’s how it works: if you come across an interesting, unusual, or potentially useful result, you then have to subject it to more and more stringent tests under different conditions. Something real will be able to pass some of these. You’ll get information and guidance from seeing when the signal is there and when it isn’t. But there are huge numbers of flickering, fluttering, hazy results that just seem to evaporate when you put the magnifying glass on them, though. Every time you zoom in, every time you give them the tiniest hurdle to jump over, they vanish. That is generally the sign of something that just wasn’t real to begin with, and the more different conditions that something fails to rise to, the more times it fails to show up, the greater the chances that it was a mistake or an illusion all along. This happens all the time in science, and in biomedical research in particular. You get used to it as a researcher. If you take every result at face value (particularly every startling result) and run with it before seeing how robust it is, you will spend your career chasing shadows. And boy howdy, are there ever a lot of shadows.
So to the people who keep insisting that one therapy or another is clearly, obviously working and why isn’t everyone using it, what’s wrong with all you people. . .well, the problem is that not many folks who practice medicine or who do medical research actually think that anything is clearly working yet at all. The data detailed in this post today are why they think that.