That’s a huge number of people, considering that nobody really knows whether convalescent plasma actually works against Covid-19.
A spontaneously generated, self-assembling group of clinicians and cross-disciplinary researchers that built the nationwide program to ensure “expanded access” to convalescent plasma also created protocols for randomized, controlled trials, the gold standard for evidence in science. They hoped to test plasma’s ability to prevent disease after exposure, its capacity to treat Covid-19—and what Michael Joyner, an exercise physiologist at the Mayo Clinic who was instrumental in setting up the expanded-access network, called a “Hail Mary” protocol to try to help people who are severely ill, on ventilators. The distribution system got approved and built; the trial protocols did not. They never began. There are plenty of reasons to think plasma might help fight Covid-19. Physicians have used it for more than a century; it’s made by taking blood from people who’ve recovered from a disease and spinning it in a centrifuge down to a frothy, yellow liquid that contains the sum total of the donor’s immune response—molecules that attack all invading germs, and some that specifically target all the individual pathogens the donor has ever encountered. But actual rigorous trials of the stuff are rare. Dozens of randomized, controlled clinical trials are underway—tests that systematically compare the same kinds of people at similar stages of the disease who get convalescent plasma to those who don’t. Even without that rigor, this year tens of thousands of people received plasma for Covid-19. It played out as a one-on-one decision between physicians and patients, not a population-scale experiment designed to elicit knowledge about its efficacy. A preprint from the expanded-access group, not yet peer-reviewed, recounts the outcomes of more than 35,000 of these recipients at hundreds of hospitals. It retroactively splits that population into groups based on when in their illness they got plasma, or how laden the plasma was with the antibodies that actually do the disease-fighting. But, as the researchers and outside experts both acknowledge, that’s not as good as a clinical trial in which people get randomly assigned to a group that gets a drug (or procedure or surgery or whatever) versus a group that doesn’t, and then someone compares the results. And that’s a palpable loss. “Fifty thousand people have been given a treatment, and we cannot know whether it worked or not,” says Martin Landray, one of the leaders of the Randomised Evaluation of Covid-19 Therapies (or Recovery) Trial in England, a large-scale, multi-center, multi-drug randomized controlled trial that showed that the corticosteroid dexamethasone saved the lives of Covid-19 patients and the autoimmune drug hydroxychloroquine did not. (That 50,000 number was from a few weeks back, just after the plasma preprint came out.) “You wouldn’t need to randomize 50,000 patients. You wouldn’t need to randomize 5,000 patients to get the answer,” he says. “But that one difference is the difference between the effort being worthwhile or not.” Here’s another perspective, using the more up-to-date number: “In my mind, treating 98,000 people with plasma and not having conclusive data if it worked is problematic, and we should have a more robust data set before we give 98,000 people a product,” says John Beigel, associate director for clinical research at the National Institute of Allergy and Infectious Diseases’ Division of Microbiology and Infectious Diseases. Beigel was the lead author on the study of the drug remdesivir that led to its incorporation into the US standard of care for Covid-19.
What looked at the beginning of the pandemic like a rare bright spot—that a classic treatment used in pandemics for the past century might prove effective in this one, too, and provide a relatively simple stopgap before monoclonal antibodies and vaccines—now looks dimmer. It’s not that it doesn’t work. It’s worse than that: Because of failures in the system of how science gets done, nobody knows.
This isn’t how things were supposed to happen. Earlier this year, when I spoke to Joyner, he told me about the three trials they hoped to run. Back in March, Joyner saw work validating plasma’s use as a bridge to more specific therapies—hyperimmune globulin, monoclonal antibodies that attack a carefully chosen part of the virus, and eventually vaccines that build a bodywide response, ideally a permanent one.
It made sense. Plasma has been used for decades. Health care workers tried it against MERS, against the first SARS, against Ebola. More recently, a small randomized clinical trial of plasma on Covid-19 in China published in JAMA in June—just 103 patients—stopped early but showed indications of improvement.
Another small RCT in the Netherlands, published in July as a preprint, didn’t show any efficacy of plasma, but it hinted that timing the dose was the key.
Mayo spokespeople declined to make Joyner available for an interview this week, but he has said on Twitter that funding sources turned his group’s trials down, that they were more interested in pursuing hyperimmune globulin, a more specific type of blood-derived immunotherapy. Another researcher, who was involved in the construction of the expanded-access program, likewise tells me that they met resistance to their efforts to run randomized trials from the start. “We drew up these trials, and we approached multiple federal agencies and private funders to immediately get these trials going, and we actually didn’t have a lot of luck with that,” says Jeffrey Henderson, a physician and infectious disease researcher at Washington University St. Louis. “We had all these trials in the hopper. We were ready to ride the first wave. We figured, we’re not going to have enough to give to anybody anyway, let’s run trials. We just could not get traction. The studies we’re coming out with now are not the studies we wanted to do. It’s people making the best of the situation.” I asked representatives of the National Institutes of Health whether they discussed the studies Joyner proposed, but they haven’t responded. Henderson told me that the Mayo-led consortium approached the Gates Foundation, a major funder of medical research, and were similarly rebuffed. A Gates Foundation spokesperson says the foundation never received a formal proposal of any kind from Joyner’s group—and that the foundation is more focused on funding research into hyperimmune globulin. It doesn’t require a relatively complex infusion, as plasma does, it’s more shelf-stable, and might be more targeted as a treatment. The expanded-access program might have been, in one sense, a victim of its own success. Its planners initially expected to reach perhaps 300 hospitals. They instead ended up with a network of more than 2,700, with 14,000 physicians. The team expected to get 5,000 patients signed up, a mark they passed within a few weeks. The vast majority were at hospitals that had no infrastructure or experience with clinical trials, and wouldn’t be expected to run them. At the beginning, the FDA expected a smaller-scale expanded-access program that would be one of three ways people would get convalescent plasma. The other two would be either through an emergency Investigational New Drug (or IND) license that the agency also approved (which would let physicians prescribe plasma they got from other sources), and randomized clinical trials. Patients would get the help they desperately needed, and the proof would come alongside soon after. The expanded-use program quickly swamped the other two routes. In that sense, the FDA’s fast and compassionate approval of the expanded-use program might also have wounded the trials. “We were pushing, pushing to get the IND to do studies, assuming that studies would be funded somewhere, and the FDA is moving unusually fast,” Henderson says. “And then what they release is a compassionate use IND. I just remember it was like, silence. Like, wait a minute—is this good? It wasn’t what we were pushing for.” On the other hand, R. Scott Wright, director of the Human Research Protection Program at Mayo and another of the expanded-access program’s coordinators, says setting up trials would have been much more difficult than people have made it seem. In March and April, the Mayo-based group didn’t have enough plasma to send to potential trial sites, nor did it have enough placebo—it didn’t even know what an appropriate placebo would have been. (Saline is clear and looks different from plasma, though you can add food coloring. But then to keep the study blinded, you also have to disguise that it comes from the pharmacy and not the blood bank.) Also, one of the hallmarks of the Covid-19 pandemic is the speed with which it moves from community to community, which means that by the time some trials got started, they didn’t have any patients to enroll. Travel restrictions meant that even if principal investigators knew where to go to get ahead of the disease, they weren’t allowed to go there to train site leads and study coordinators. Even if all of that got figured out, patients were skeptical of signing up for randomized trials of some complicated gunk they’d never heard of. Those who wanted plasma could already get it under the expanded-use program. Why risk getting randomized to the control group and getting a placebo? And as a Wednesday statement from the Mayo Clinic put it, the expanded-access program was “not intended to be a clinical trial to determine the efficacy of convalescent plasma,” even though it did convey “possible signals of efficacy.” The expanded-access program has involved the time and energy of thousands of physicians and nearly 100,000 sick people. Yet it has generated little information to help the next set of physicians and patients make a decision about whether to use that same compound. The resulting preprint isn’t nothing, to be sure. It shows that convalescent plasma is safe to use, as it has been in pandemics past. The researchers were able to sort their data to extract some knowledge. They kept track of the timing of when plasma was administered—early after a patient showed symptoms, or later. Earlier use seems to translate to lower mortality, according to the preprint. A reliable test of the levels of antibodies in the plasma—called antibody titer—didn’t exist when the program started, but the researchers made sure that all the blood banks involved saved samples of the plasma that got administered. Later, when tests did exist, the researchers went back and determined whether plasma given to someone contained high, medium, or low amounts of the various immunological components. So they grouped patient data that way, as well. Higher-titer plasma was most effective. “There were elements within this pragmatic design that allowed for something similar to randomization. I call it pseudo-randomization,” says Wright. He’s also an author, with Joyner and many others, on the preprint. “I’d like to see a randomized trial designed to test the hypothesis that our data reveals. And, in the absence of a randomized trial, I believe the data support the earlier use of convalescent plasma, and plasma with more evidence of immune activation.” But since none of those outcomes were randomized against control groups, it’s hard to conclude anything … conclusive, if you see what I mean. Coaxing useful results from a large-scale trial isn’t impossible. Landray’s Recovery trial has done it. It’s a multi-armed study that began with tests of half a dozen drugs, including the steroid dexamethasone (success! and now part of the Covid-19 pharmacopeia) and the antiinflammatory hydroxychloroquine (bzzzt). Recovery even has an arm studying convalescent plasma, though the decline of cases in the United Kingdom means enrollments have, for now, slowed.
Recovery owes some of its success at putting drugs through a scientific wringer to the UK’s National Health Service. Every hospital is linked to the same database, with the same interchangeable record format. It’s easy to randomize new patients and collect data on them.
That’s the kind of thing that has convinced some trialists in the US to advocate for a standing “pandemic protocol,” a pre-prepared network of research hospitals ready to mount trials of whatever, as soon as a pandemic hits. It’d be hard—nobody likes to spend money for no apparent reason, just waiting around for disaster. (Though in fact such a network could do other equally useful work during pandemic halftimes, and might well save money and lives in the long run.)
Still, it’s hard not to be frustrated by what looks like a lost opportunity this time around. Tens of thousands of people got convalescent plasma, and nobody really knows if it helped them. Nobody really knows if it’ll help anyone else. “Future patients around the world, each one that comes along, you make an arbitrary decision whether to give it or not based on no more information than the last patient. You never learn anything,” Landray says.
The head of the NIH, Francis Collins, convened a meeting a couple of weeks ago of many of the people receiving agency funding to do randomized, controlled trials on convalescent plasma. Half a dozen researchers, including Joyner, presented their data via videoconference. It might not have been a turning point, but it does suggest renewed interest in coordination of trials.
Just a few weeks ago, a New York-based trial led by Liise-anne Pirofski, the head of infectious disease at Montefiore Hospital and a professor at Albert Einstein College of Medicine, was stalled at two hospitals (NYU was the other) and only 180 enrolled subjects. Pirofski had put the trial, her first as principal investigator, together in April while she was also treating New York’s surge of Covid-19 patients. “I spent quite a bit of time calling people at sites that were surging, and to be honest I didn’t get a lot of buy-in,” Pirofski says. But now she has $4.5 million in additional NIH funding under “Operation Warp Speed” and has expanded to research centers in Florida and Texas. “When we got funding and we had conversations with NIH, they saw the importance of expanding the trial and really helped us,” she says.
Her trial is a straight, head-to-head comparison of hospitalized patients randomized to get transfused with either plasma or a placebo made of saline solution. An adaptive approach to the data means that even though she hopes to get at least 300 participants, an independent data team will monitor the ongoing results. If a signal shows up sooner, they’ll let Pirofski know. Results could be a kind of closing-of-the-loop; with Arturo Casadevall of Johns Hopkins, Pirofski wrote one of the early, influential articles advocating for the use of convalescent plasma against Covid-19. “What I really love about our study is that we will get an answer, and the answer will be the following: Plasma works, or plasma doesn’t work,” she says. Meanwhile at Hopkins, physician and infectious disease researcher Shmeul Shoham is running one of two companion studies that echo the protocols the expanded-access folks wanted to try—a post-exposure prophylaxis study of people exposed but not yet ill, and another of people who have symptoms but haven’t yet been hospitalized. At first, his money came from the Bloomberg Foundation and the state of Maryland. Then the Department of the Defense came through with $35 million. They’ll have over 1,000 participants at as many as 30 sites across the US, including in the Navajo Nation. “I have not had a good night’s sleep since March—worrying about whether we’re going to get approval, and then funding, and then enrolling patients,” Shoham says. That’s just a smattering of the trials that are out there. A team at NYU is trying to use smart statistics to combine, meta-analytically, the data that the small and paused plasma trials have gotten, in an effort to squeeze some useful knowledge out of them. And while it’s tempting to bang on a table about the lost months when thousands of people got plasma without a rigorous evidence base, it’s also true that lots of standard medical practice doesn’t have that evidence—for many of the same reasons of cost, time, tradition, and so on. (Consider medical interventions for low back pain, or the ways oncologists review imaging test results, or nutritional science, or or or.) One truism about randomized controlled clinical trials is that they are expensive. Another is that they take forever. And even though it might feel otherwise, Covid-19 is only eight months old. “Having done clinical trials now for 20 years, the time between when an idea is thought of, and the time to when the study is funded and operationalized, is usually a couple of years,” Shoham says. “And then the time from when the study results come in to when it changes therapy in a meaningful way can be another five years after that.” The bungled Covid-19 response in the US doesn’t allow that kind of time. Sure, randomized, controlled trials won’t always be possible, or even appropriate. Are they the only way to know about the world? No, of course not. Pragmatic studies like the expanded-use preprint or retrospective and observational studies are all stations on a journey to greater (but never perfect) certainty. “I’m a trialist. I strongly advocate using trials when appropriate. But I think there are other ways to get medical evidence that may not be as powerful as a trial but are important,” Wright says. That’s manifestly so. But in the case of convalescent plasma, a therapy that seemed like a tantalizing possibility in March, remains only that: tantalizing.
