There was a bit of news on each of three coronavirus treatment fronts this week: hydroxychloroquine, convalescent plasma, and Remdesivir.
At the end of April, the NIH announced that Remdesivir helps people hospitalized with COVID-19 recover a bit faster. They didn’t actually share many details in their press release. Now, the peer-reviewed paper with more information is out. The paper has additional data breaking patients down by severity – whether they needed oxygen or a ventilator, for example. These subgroups get pretty small pretty quickly, though, so it’s hard to reach any strong conclusions about how the effects of the drug differ between these groups.
Remdesivir is the only treatment to date with support from high-quality randomized controlled trials. The studies I’ll discuss below for other drugs weren’t randomized, so they can’t tell us as much either way.
[6/4/20 Update: The paper discussed in this section was retracted today. The research relied on a commercial database. The company that owns that database says it is unable to share information that would be needed to run an audit of the data because it would violate confidentiality agreements. See the full statement from The Lancet here.]
This new paper in The Lancet collected medical records for 96,000 people hospitalized with COVID-19. They broke patients into five groups: a control group and four treatment groups that received chloroquine or hydroxychloroquine and did or didn’t receive another drug that’s often being paired with those. They were able to get baseline health information for each patient, as well as a basic measure of how severe their COVID-19 case was. With that information they can attempt to adjust for selection bias; you’d naturally worry that patients are given these other drugs only if their cases are more severe, which could make the drug look harmful even if it’s helpful. Still, without a randomized trial, we can’t be sure this problem is solved. The study measured two outcomes: death and an abnormal heartbeat, which is a known side effect of (hydroxy)cholorquine. All four treatment groups had a statistically significant increase in mortality compared to the control group, about a 20%-50% increase across the confidence intervals for all four treatments. And all four treatments had an increase in abnormal heartbeats, ranging from about 2 times to 6 times the rate in the control group. At the high end, that’s an even higher risk of abnormal heartbeat than people who had a history of abnormal heartbeats before they got COVID-19. Because this study wasn’t randomized, we shouldn’t treat it as the final word on hydroxychloroquine as a treatment for COVID-19. But these results look pretty bad, so people should only take these drugs as part of a clinical trial and with eyes wide open that the benefits are uncertain and at the expense of increased risk of abnormal heartbeat.
Two weeks ago, researchers from the Mayo Clinic, Michigan State, and Johns Hopkins released a study of convalescent plasma (blood transfusions from people who have recovered from COVID-19) in 5,000 patients. These patients weren’t randomized to receive the treatment or not. They were given convalescent plasma specifically because they had severe cases of COVID-19. That makes it difficult to infer how effective the treatment was. This study doesn’t even try to do that, it just checks whether convalescent plasma is safe.
Of the 5,000 patients, most of whom were in the ICU, 15 (0.3%) died within four hours of their transfusion. Within a week, 15% of the patients had died. These mortality rates are roughly similar to rates previously published for patients hospitalized with severe COVID-19, so the authors judge that convalescent plasma transfusions are probably safe. Needless to say, this is an extremely rough measure of safety.
This week, researchers from the Icahn School of Medicine at Mount Sinai released a small, non-randomized trial that used similar patients as a control group. The treatment group included 39 patients. The control group was made up of patients at the same hospital. Control patients were paired with treatment patients based on 1) treatment with hydroxychloroquine, 2) intubation, 3) length of hospital stay, and 4) oxygen requirement.
They tracked two outcomes: oxygen requirements and death. 82% of the treated patients improved on the scale of oxygen requirements, compared to 76% in the control group. Although that difference is small, it did reach statistical significance. 13% of the treatment group died, versus 24% in the control group. When they split the groups into patients who were or weren’t intubated, they find that the non-intubated group had an 80% reduction in death if they received convalescent plasma but the intubated group had a (statistically insignificant) 24% increase in death. Keep in mind that once you split these 39 people into two groups, you’re left with an absolutely tiny, non-randomly selected treatment. We should only give these results a little bit of weight in updating our beliefs about how effective convalescent plasma is. With those caveats, these results do look very good.