We’ve had several releases of remdesivir data, and this was not exactly one of those controlled-release formulations. No, this was more in the “chaotic mess” category, with news items coming from several sources with partial information. I decided not to blog on it until the loud banging noises stopped, and I’m glad I did, since I would have had to have revised (several times) whatever piece I managed to crank out.
Recall the previous state of affairs: we had a readout in 53 patients as reported by Gilead in the NEJM (of all places – future readers will be able to tell that they’ve hit the 2020 stratum in the literature when they run into a bunch of papers on clinical virology that all seem like they’re in the wrong journals). That one was interpreted as somewhat positive, although there was really no way of knowing. Then there was the news that a remdesivir trial had been suspended in China, which took everyone by surprise and did not breed confidence in anything: the drug’s prospects, the results themselves, the Chinese, you name it. A day or so later came word of leaked data from a trial in Chicago that sounded much more positive, and that threw some people’s opinions back the other way.
Now we have bits of data from an NIAID/NIH trial of the drug that began enrolling in Nebraska in late February (the first patient was one of the Diamond Princess cruise ship passengers, which seems long enough ago by now that it might as well be the Titanic). This one was double-blinded and placebo-controlled: patients in the treatment group received 200mg of the drug the first day and 100mg each day thereafter, for up to ten days. Participants needed to test positive for the virus and have evidence of lung involvement in the disease. The primary endpoint was improved time to recovery (discharge from the hospital or ability to return to normal activity), and it appears that remdesivir was statistically better than placebo: 11 days versus 15 days. The team also monitored overall survival in the >1000 patients, and there was a possible trend towards the drug, but it did not reach statistical significance (8% mortality in the treatment group, 11.6% in the placebo group).
That’s it. Those are the numbers we have. The rest will be in a “forthcoming report”, and we’re going to have to wait until it comes forth. This release was after an April 27 meeting of the data and safety monitoring board, and it’s worth noting that had there been “clear and substantial evidence of a treatment difference” during the trial that the DSMB was to have halted the study at that point. We can infer that nothing rose to that level, then: we have a difference, but not substantial enough to have ended the trial prematurely. And I have to note another issue: if you look at the Clinicaltrials.gov record for the trial, it appears that the outcomes measures for the trial were changed (as noted by Walid Gellad on Twitter). That primary endpoint of the trial mentioned above, time to recovery? It was originally an 8-point severity scale (death, on ventilator, hospitalized with oxygen, all the way down to discharged with no limits on activity). A similar ordinal scale measure is still in the secondary endpoints, as it was before, but we have no numbers for that yet, of course. But it’s clear that the primary endpoint was changed at some point in April.
The other piece of news is this paper in The Lancet describing a randomized, double-blind, placebo-controlled trial in China with 237 patients – in fact, the very one that was halted suddenly. This one enrolled patients that had 12 days or fewer of symptoms, confirmed viral infection, and involvement of the lungs as well (<94% oxygen saturation on room air and confirmed viral pneumonia on X-ray). The dosing of remdesivir was the same as in the NIH trial. But in this one, the use of the drug was not associated with a shorter time to clinical improvement. A subgroup analysis showed a trend (not statistically significant) towards shorter duration in the patients who overall showed symptoms for ten days or less, though. Mortality was identical between the two groups, although there was again a trend (not significant) towards less mortality on remdesivir in the shorter-duration patients, and thus a trend towards higher mortality in the others. The viral load was checked in both the upper and lower respiratory tracts of the patients, and remdesivir had no effect whatsoever on it compared to placebo, in any group.
So that paper’s results are not very impressive, frankly, and if you’d seen that one first then the NIAID results would have been even more surprising. I very much look forward to seeing the viral load numbers from that one, since that was one of the more striking misses in the Lancet paper. Remember, the mechanism of action of remdesivir, as a nucleoside mimic, is to interfere with viral replication. So if viral load isn’t be affected, you have to wonder what’s going on.
It’s possible that the drug’s efficacy, such as it is, is most apparent when the drug is given earlier in the disease course or to patients with are getting severe disease in general (those two categories probably have substantial overlap, and early in the infection you probably can’t tell the difference). Many people yesterday were saying that sure, the Chinese patients in the Lancet paper were in worse shape than the ones in the NIAID study, but looking at the inclusion criteria, I really don’t think that they were that different. The authors of this study note that their patient population was actually less ill than the ones reported in Gilead’s earlier NEJM paper, and that they had hoped, based on this, to have seen more of an effect than they did.
Taken together, the picture that’s emerging is that remdesivir may be of some help in less-severe cases. It is not a cure; a cure would have shown up in the trials we’ve run already, and cures are mighty thin on the ground for viral diseases. We can hope that the time-to-recovery is actually a useful measure and that the drug might get people out of hospitals a bit earlier, and hope a bit harder that there really is a mortality difference that will turn out to be real as we go into larger and larger numbers of patients. But working against that is the possibility that wider use of the drug will obscure the effect rather than make it more obvious.
To forestall some questions that I know will come up: what do I think about this versus hydroxychloroquine? Well, we have more controlled data to work with on remdesivir, for one thing, so whatever benefits there are, are more obvious. The balance of what controlled data we have on HCQ is negative, and here we’re at least more mixed. There is also (to the best of my knowledge) no particular safety signal for remdesivir, as opposed to HCQ (particularly the HCQ/azithromycin combination). So while I’m not bowled over, I’m more optimistic than I am about hydroxychloroquine.
That said, I continue to think that we do not yet have any treatment regimes that are making a big difference, unfortunately. I saw a tweet from Nate Silver, no fool, wondering about the possible effects of having some combination of drugs by this fall, things that could be widely administered so that anyone who needs them could get them, that reduces mortality by a significant amount (he mentioned 30%, 50%, or more). He started out by saying “Non-rhetorical question”, so taking that non-rhetorically, all I can say is that I don’t see it happening. Speculation about it seems a waste of mental effort. Remdesivir was honestly one of the better shots, and at best it might have decreased mortality from about 11% to about 8%. I don’t see any combination of repurposed drugs adding up to the numbers that Silver is wondering about. The only new (non-repurposed) therapy that could do it in that time frame is, I think, a monoclonal antibody. Having that widely available by the fall would really be pushing it, since it looks like the first patients will be dosed in June for the very fastest-running efforts. “Widely available” is the key phrase: I certainly think an effective mAb can be found, and the number of companies working on finding one makes it more likely than ever that we’ll get a good one. But proving that it’s good and (most importantly) getting it produced and distributed – well, “fall” to me starts in September, and that would be quite an accomplishment. Note that I have not even talked about cost. (This is a good point to mention that even remdesivir production is not a straightforward process, either).
Until we have such a mAb, and until the advent of a vaccine later on, I do not see any game-changers on the horizon. I will look forward to being wrong about how quickly these things will appear – that would be great – but I don’t think I’m wrong about those being the main things that will knock down the virus.