The much-hyped drug sparked a battle between power and knowledge. Let’s not repeat it.
In the mid-1600s, a Jesuit priest serving in Peru got a useful tip. The indigenous people there, he learned, were using the bark of a particular kind of tree to treat fevers. The priest, who’d probably gone a few rounds himself with the local diseases, got ahold of some of the reddish-brown bark from this “fever-tree” and shipped it back to Europe. In the 1670s, what came to be called Jesuit bark had made its way into a popular patent medicine, along with rose leaves, lemon juice, and wine.
That was the beginning of the impressively effective bark’s role in pharmacology (and its side career in mixology). In the mid-1700s the prolific Swedish taxonomist Carl Linnaeus gave the tree’s genus its name—having heard a fanciful (and untrue) tale about the bark’s success treating the Spanish Countess of Chinchón, he dubbed it Cinchona. In 1820, French chemists isolated the active ingredient, a plant alkaloid they named quinine. Its bitter flavor became not only a hallmark of the prevention and treatment of malaria but also the basis for a medicinal fizzy water—a “tonic”—that mixed well with the gin that Europeans brought with them to their equatorial conquests. Today quinine can be found in bitters, vermouth, and absinthe; next time you order a Manhattan or a Sazerac, give a little l’chaim to the Peruvians.
Medicine that treats a deadly disease but grows only on certain finicky trees is the kind of thing chemists live for. A failed attempt to synthesize quinine in the 1800s had accidentally produced the first synthetic pigment (a lovely shade of mauve); after World War I, when endemic malaria arguably did almost as much as Allied soldiers to limit Germany’s expansionist ambitions, that country set its scientists to solving a problem. A dye company called Bayer took up the quinine challenge, synthesized some reasonably useful replacements, and became a pharmaceutical powerhouse with a global market. When World War II denied the US access to both German drugs and the quinine-producing cinchona trees of Java, the Americans basically stole a recipe from German prisoners of war and turned that into a successful treatment.
That drug was called chloroquine. It has a slightly better-tolerated cousin, hydroxychloroquine. You may have heard of them.
So, yeah: A drug extracted from indigenous knowledge to lubricate European colonialist impulses went on to power the military adventures of the latter 20th century and save millions of lives. But even as the parasites that cause some strains of malaria began to develop resistance to chloroquine, newer science started to hint at a second life for the drug. Some lab studies suggested that it could fight viruses, and that it could suppress overreactions by the human immune system. By the mid-1950s, doctors were using hydroxychloroquine to treat the autoimmune disorders lupus and rheumatoid arthritis. The drug was readily available. It had manageable side effects. And because it’s so old, no pharmaceutical company holds a patent on it. So it’s cheap.
Viable. Safe. Available. Inexpensive. What more could you ask for?
It made sense, then, that when a novel coronavirus appeared in Wuhan, China, in December 2019, people started speculating about the old drug. Chloroquine’s virus-fighting reputation preceded it. Four centuries of the history of science came crashing into the newly apocalyptic present. By February, several Chinese research teams had spun up small trials of chloroquine and hydroxychloroquine against the new disease, and some were soon reporting success. A simple, familiar drug was offering hope.
Still, though. Before you start giving a drug to the thousands, soon to be millions, of people affected by a pandemic virus, you want to be very, very sure it’s safe and effective, that the benefits of administering it outweigh the risks and side effects. The Chinese studies of chloroquine were, so far, preliminary and small-bore. And because of language barriers, limited access to international journals, and some mutual distrust, Chinese data doesn’t always make it into the global information ecosystem. Nobody really knew, authoritatively, if the drug actually worked.
But “Does it work?” is a harder question to answer than it sounds. Few drugs are penicillin-size successes; most drugs have more moderate effects. That means those possible effects are hard to distinguish from what may just be statistical noise. Under normal conditions, distinguishing one from the other requires painstaking, time-consuming research protocols and statistical analysis. But the urgency of a pandemic makes conditions abnormal in the extreme. Faced with intensive care units full of the severely ill, physicians begin to feel they can’t wait for statistics before their patients become one. Politicians start looking for a win, or something to signal they’re dealing with the problem. And the world’s technical and economic elite start looking for quick fixes and opportunities to make a sale, spreading their opinions (whether quarter-, half-, or fully baked) on social media. After all, influencer and influenza share the same etymological root.
At issue here is more than just whether a drug treats a disease. The heart of the scientific method is the process of formulating a hypothesis and collecting data to test it. This is how to reliably be sure that (in this case) a drug does what you say it does—that the effects you think you see are not coincidence or luck or mirage. It sounds simple, but in practice it’s ambiguous, messy, and often contentious. The twisted tale of hydroxychloroquine is actually about how to know stuff, the question that has defined every existential decision since the early 20th century—climate change, vaccines, economic policy. We’ve learned from failure and bitter experience that only when we take the time to find the truth do we at least have a chance to make good decisions. We also know that it’ll be a struggle—that grifters, power-seekers, and fantasists will push their own versions of truth while scientists and policymakers grapple with the lumbering process and nuanced outcomes of the scientific method. Because there will be other pandemics, other disasters. And just as with Covid-19, only science and its tools will soften their impact. But also as with Covid-19, humans will do that science and wield those tools, and that makes things messy. What happened with hydroxychloroquine was a debacle, but retelling the story might help avert the same kind of chaos next time around.
Viruses aren’t alive, exactly—they’re just genetic material wrapped in fat, starch, and protein. But because they use living things to reproduce and spread, evolutionary forces effectively shape them, synchronizing viruses with the specifics of their targets. Viruses land on cells, and viruses’ landing gear, as it were, are shaped to lock onto the exact topologies of proteins on their surfaces. Once clicked onto that docking site, a virus forces the cell to engulf it in a little bit of membrane. Like a fighter jet on an aircraft carrier deck, the virus gets elevatored into the cells’ innards. Down there, the viral genes slide into the cell’s own genome and take over, forcing the cell to pump out more copies of the virus. Eventually the cell bursts open, the new virus copies spread, and the process starts all over.
Hypothetically, chloroquine and hydroxychloroquine can mess all that up. They interfere with the biochemistry that lets the landing gear touch down, a process called glycosylation. And it seems like the drugs change the acidity of the elevator shaft, of that bit of involuted membrane bubble, making it inhospitable to a virus and preventing infection.
It works in the lab, anyway. Over decades, researchers have tried chloroquine and hydroxychloroquine against a bunch of viruses, including the human immunodeficiency virus that causes AIDS. The new pathogen that emerged in 2019, SARS-CoV-2, belongs to a family called coronaviruses—as did its prequel, SARS-CoV, which caused severe acute respiratory syndrome. In 2004, a team of Belgian researchers tried chloroquine on SARS-1 in the lab, and it seemed successful—apply the drug to cells and the virus has trouble replicating.
Cells in a petri dish aren’t people, but even with such crappy evidence, it made sense in the early days of the pandemic to try the drug again. Emergency rooms and intensive care units were filling up with sick people who couldn’t breathe, and frankly, frontline caregivers didn’t have much else to give them.
By March 9, the US was facing a shortage of hydroxychloroquine and chloroquine. About a week later, with a surge of Covid-19 patients slamming New York City, I talked to Liise-anne Pirofski, the chief of the Division of Infectious Diseases at Montefiore Medical Center and the Albert Einstein College of Medicine. Chloroquine was standard for patients with Covid-19, along with a repurposed HIV antiviral—even though, at the time, there was only the thinnest data recommending either drug. “Everybody gets that unless they have some contraindication,” Pirofski told me. What else could they do? Her hospital was participating in a clinical trial of a then-experimental antiviral called remdesivir, but it was still unavailable outside that study. Pirofski herself was advocating the use of convalescent plasma, a decades-old treatment made from the blood of people who’ve recovered from a disease, which also hadn’t been tested against Covid-19. They were throwing everything they had at the virus. People were sick and dying. You go to war with the drugs you have, not the drugs you wish you had.
The possibilities in early 2020: Hydroxychloroquine might help. Or it might not. Or it might make people worse. No one knew.
One of the first people to leap into that breach was David Boulware, a diligent infectious disease researcher and professor of medicine at the University of Minnesota. Back in 2015 he’d worked on an Ebola drug trial with the National Institutes of Health, and he quickly raised his hand to work on trials of treatments for the new virus.
In early March, he and his team were supposed to be at an HIV conference in Boston, but by that point nobody was traveling anywhere. “We all had four days free to totally focus on this task,” Boulware told me then. His group used the time to put together a plan to study hydroxychloroquine.
Right here—the stage where scientists come up with these “research protocols”—is where how-to-know starts getting complicated. It’s a cliché because it’s true: The answers you get depend on the questions you ask. In this case, Boulware’s team decided not to test the drug on hospitalized patients, when the disease becomes severe. “If it was going to work, you’d have a better chance to alter the disease course early on,” Boulware said.
They hoped it worked. But they didn’t know. To find out, they proposed a classic structure: A couple hundred people would get the drug; a similar number would get a placebo—an inert fake. The ones getting the placebo would be the “control group,” experiencing all the same things except for the drug, to isolate its effects. Neither researchers nor participants would know who got which until the end; that’s called a “double-blind” study. And people would be assigned to the groups at random, to avoid even unconscious bias on the part of the researchers and prevent differences between groups of humans—socioeconomic, demographic, and so on—from throwing off the results.
That is, in other words, a large, double-blinded, randomized controlled trial. Boulware’s team proposed two. One would look at whether hydroxychloroquine could prevent illness in people with exposure to an infected person—“post-exposure prophylaxis”—and another would see if taking the drug close to the onset of symptoms could keep those symptoms from getting worse. That was “early treatment.” On March 13, the US Food and Drug Administration approved the study, a blisteringly fast green light from a typically cautious, plodding agency. The responses of the federal government’s scientific policymaking would falter in key ways over the next few months, but this wasn’t one of them.
Boulware started enrolling people almost immediately. For statistical validity, they’d need enough people so that some in the experimental groups and some in the controls would get Covid-19. The researchers would run the numbers, ask who got what, and they’d have an answer in weeks. They’d write up the results, publish in a journal, and it would be science.
Except Boulware’s reasonable expectation that things would work the way they were supposed to didn’t take into account the viral social-media blender that was spinning up its blades—making a viscous gazpacho out of Silicon Valley opportunism and the hottest of hot takes from the president of the United States.
The way they were supposed to? Yeah, no.
Even the stodgiest of scientists don’t believe that waiting months or years for a formal write-up of an experiment to penetrate a wall of skeptical reviewers, receiving an inscrutable thumbs-up to get published—in ink! on paper! that gets mailed! to libraries!—is an ideal system for disseminating new knowledge today. Yet that’s still mostly how things work, despite the existence of the online version of most journals. But the Covid-19 pandemic came at a weird moment in the history of how information spreads. For one thing, that formal system was already in the process of breaking down. Due to the pressures of publication and academic seniority, some of the science that gets into peer-reviewed journals doesn’t hold up to scrutiny, and many scientists are internalizing the hard truth of that “ reproducibility crisis.” Formal peer review and publication doesn’t make something true. That’s part of the reason the biomedical sciences were embracing a newer approach, one that their colleagues across the quad in the physics and math buildings had arrived at years before: “prepublication” or “prepress” articles that could go online as soon as their authors finished typing them.
That’s good; it means faster, freer information and a more egalitarian kind of review. But rethinking the gatekeeping in the ways nominal experts disseminated nominal knowledge opened the door to other people playing the game. Thanks to widespread access to publishing tools and social media, pretty much anyone can marshal the trappings of expertise. The crisis of the global pandemic intersected with a crisis of belief, with opposing scientific ideas somehow getting tethered to political ideologies. With just a bit of Googling, anyone can find things that look like truth, that are what that person was hoping to hear in the first place. If one of those things goes viral, and if the science behind it is difficult or undercooked, pretty soon everyone starts nodding along.
Which is what happened on March 13—the same day the FDA approved Boulware’s well-thought-out trial. A physician named James Todaro tweeted that chloroquine could fight Covid-19, and he’d written a paper that proved it. Now, this wasn’t a “paper” from a peer-reviewed journal, or even a preprint. It was a Google Doc, coauthored by a lawyer named Gregory Rigano and a biochemist named Thomas Broker, identified as a Stanford PhD. It was a pretty good summary of all the research on chloroquine up to that point. It even cited the work of a French researcher named Didier Raoult, a controversial infectious disease specialist who, a few days later, claimed he had results showing that hydroxychloroquine worked against Covid-19 in human beings.
A steady rain of likes and retweets turned into a viral downpour. The influential Silicon Valley blog Stratechery linked to the Google Doc. Rigano went on Fox News. Elon Musk tweeted about the document with the link. Musk, who said he’d taken chloroquine for malaria, also tweeted a link to a video on hydroxychloroquine and Covid-19 produced by a small medical-education company called MedCram. The company had started doing brisk traffic covering the coronavirus; the hydroxychloroquine episode took off.
The original Google Doc made a good case for chloroquine being of interest—attempted use in prior pandemics, studies in cells and in animals, preliminary results from China. Not proof, to be sure, but tantalizing hints. But, as it turned out, the creators were not all that they appeared.
Rigano had done most of the initial work. According to his LinkedIn bio, Rigano was on leave from a master’s program in bioinformatics at Johns Hopkins and was an adviser to a drug development program at Stanford. But the head of the bioinformatics program at Johns Hopkins told me Rigano wasn’t really on leave from the program; he had only taken one class. And the codirector of the Stanford program told me that, while he’d met Rigano, he was in no way an “adviser.” Todaro, whom Rigano met via Twitter, was a former ophthalmologist turned professional bitcoin investor. And Broker was not, it turned out, a Stanford biochemist. He attended Stanford but now was a retired virologist at the University of Alabama who studied not coronaviruses but an entirely different family of viruses. Broker disavowed any involvement in the paper, and Todaro and Rigano soon removed his name from it.
None of which is to say they were necessarily wrong. But none of which is to say they were necessarily right, either. Yet the idea rippled through Silicon Valley like photons through an optical cable. Facebook, Amazon, Apple, and Google had sucked up most of the disruption oxygen in tech, and entrepreneurial types were already interested in biotech as a thing to pour money on. And their libertarian bent means they’re always looking for an institutional eyeball into which they can shove a venture-capital finger. The medical establishment, with its elitist reliance on the plodding, 20th-century model of clinical trials in the midst of a raging pandemic, seemed like a fat target.
The need for speed was real, and it played into the baser, basic instincts of the Valley. Those hold that all a technologist needs is a dream, a minimum-viable product, and the will to build a company. (A Stanford undergraduate degree doesn’t hurt.) If you’re trained to see your successes as the result of genius and instinct rather than luck, you might not be able to readily distinguish between the rigors of testing a drug’s efficacy and the travails of bringing a product to market. But they are different processes with different goals. In the Valley, whether something works is different from, maybe even disconnected from, whether it sells.
Combine that with the quantified-self, n-of-1 approach to health and wellness that some of the same people also embrace, and you get not science but pseudo-science touted by the four-hour-body crowd that gets rejuvenating transfusions of young people’s blood and rebrands nutritional diet shakes as food from a dystopian science fiction movie. “Tech, and especially Silicon Valley, has this belief that all you have to do is disrupt things and try shit and make it stick to the wall, and it will work and change everything,” says one investor with a long history in health care. “We’ve had a tried-and-true method of getting vaccinations and drugs approved in the US that is absolutely antithetical to everything the tech industry believes and has found to be true.”
“We’ve had a tried-and-true method of getting vaccinations and drugs approved in the US that is absolutely antithetical to everything the tech industry believes.”
As deaths in the US mounted and the economy went into a lockdown-induced spin, some rich and successful venture capitalists started arguing that the whole system was nonsense. As noted contrarian, investor, and former PayPal, LinkedIn, and Square executive Keith Rabois tweeted, “Randomized controls are horrible ideas. Largest impediment to progress in health spans.” (Rabois agreed to consider answering emailed questions but didn’t respond to the ones I sent.) Randomized, controlled trials not only take too long, Rabois and his ilk said, but were in this case unnecessary. You could instead use “real-world data,” like the experience of the tens of thousands of people who were actually taking hydroxychloroquine, and do some kind of data thing on it.
It’s not crazy. Randomized controlled trials are, as the scientists say, the gold standard. But that method isn’t the only way to figure out causality, or at least to start to get a sense of it. Sometimes double-blind studies are impractical. Sometimes nature and circumstance offer a great opportunity to see how changes in conditions have different effects. Observational studies, retrospective analyses of existing data, meta-analyses of grouped smaller studies—they’re all useful, and certainly better than throwing biotechnological spaghetti against a pandemic to see what sticks. But look what happened months later, after similar hopes for convalescent plasma as a therapy turned into widespread use. After giving it to nearly 100,000 people, plasma appeared to be safe, but there was only limited evidence of its effectiveness.
If it’s possible to characterize an entire swath of opinions, though, what the techfluencers seemed to be pitching was not a study where the parameters of observation were defined in advance, but one where all sorts of casually collected data, the flotsam and jetsam of our digital lives, might somehow be tabulated and correlated to whether, when, and how a person got hydroxychloroquine. Quantified self, but applied to everyone—quantified other.
To be fair, the ethics of demanding rigorous, time-consuming tests during a pandemic are worth debating. In a sense, this is about medicine now versus science later. Correctly administered, hydroxychloroquine only rarely has serious side effects; it’s a well-understood, mostly safe drug. Why not just give it to everyone and monitor their outcomes? That’s a very Silicon Valley approach—intermediate risk, high reward. “I appreciate some of the tech people coming to health care, because I do think we should be thinking about some things differently. Having fresh thinking is great. But fresh thinking is different from illogical thinking or uncaring thinking,” the investor tells me. “If you’re a tech guy flacking hydroxychloroquine to people who shouldn’t use it, what the fuck? People can get really sick.”
Even if they don’t get sick, that plan still has problems. Giving people a drug that may or may not work is ethically dicey. And who would actually keep track of those outcomes? “Big data” approaches to medicine are susceptible to the distortions and bias of anecdotal evidence and intuition, exactly the mistakes that rigorous, large-scale, randomized controlled trials are designed to avoid. But over decades, those trials have gotten more and more complicated and expensive—just as government funding of them has plateaued. The main consequence has been that pharmaceutical companies fund their own trials, and the companies are highly incentivized to focus on drugs with huge potential markets. That often means more expensive lifestyle drugs and fewer worthy public health solutions or medicines with population-scale benefits—more Viagras, fewer Vancomycins. Little wonder, then, that researchers running trials for the unpatented drug hydroxychloroquine had such trouble gaining traction, while the expensive antiviral remdesivir, with the transnational pharmaceutical company Gilead Sciences pushing it, found support for a trial in the NIH and in the White House—and is now standard in US Covid-19 treatment. The foxes all run their own chicken-coop businesses.
The same week the mania for the drug took hold in Silicon Valley, Larry Ellison, the chair of Oracle and the fifth-richest person on earth, started talking with Donald Trump. According to The Washington Post, Ellison wanted to pitch a widespread study of chloroquine and hydroxychloroquine as a treatment. Ellison proposed that Oracle could develop a website to track people’s use of the drug along with their health outcomes, and the data would anticipate whatever a slow, expensive randomized controlled trial might eventually reveal. (Through a spokesperson, Ellison declined to answer my questions about these discussions, as did a White House spokesperson.)
Ellison seemed to make an impression. Shortly after that conversation, the Post reported, Trump met with his senior advisers on the coronavirus pandemic and asked if the government could expedite the approval process for hydroxychloroquine, chloroquine, and, for good measure, remdesivir. Emergency use authorizations had been employed during pandemics in the past, to allow treatments with potential to jump the line in times of urgent need. Remdesivir was in the midst of a large-scale randomized trial sponsored by the National Institutes of Health. Hydroxychloroquine didn’t have the same backing.
The president’s urgency wasn’t just a matter of public health. Trump had promised Covid-19 would just disappear, but the US response to the disease was going entirely off the rails. During a disastrous visit to the CDC on March 6, Trump touted his own scientific acumen—“I like this stuff. I really get it. People are surprised that I understand it”—but behind the scenes he was obstructing programs to begin widespread testing for the disease. The failure to do those tests meant that as March ticked onward, thousands of Americans were already infected. Trump acknowledged privately to the journalist Bob Woodward that Covid-19 was a dangerous, plague-level disease even as he railed against the press on Twitter and elsewhere, hoping to bolster a plummeting stock market. (“I don’t want to create panic,” he said in September when asked about why he had downplayed the severity of the pandemic.) And meanwhile every model, every infectious disease researcher, every epidemiologist was looking at case and fatality curves on the cusp of exponentiality, with worst-case fatality estimates in the millions.
A miracle cure must have sounded pretty good.
On March 19, the president conducted a press conference, and it was really weird.
This is where he started pitching hydroxychloroquine. “It’s shown very encouraging—very, very encouraging early results. And we’re going to be able to make that drug available almost immediately,” the president said. The FDA was all in too: “They’ve gone through the approval process; it’s been approved.”
This was untrue in most respects. Few results were in. The president might have meant that hydroxychloroquine was approved for malaria, lupus, and rheumatoid arthritis, and that clinicians could prescribe it off-label. He might also have been talking about Boulware’s trial, which had also been approved by the FDA. It’s certainly possible the president got confused.
The president introduced FDA commissioner Stephen Hahn, who treaded cautiously. Chloroquine was worth considering for use against Covid-19, Hahn said. “Again,” he said, “we want to do that in the setting of a clinical trial—a large, pragmatic clinical trial.”
That wasn’t what the White House was pushing for behind the scenes, though. At that same moment, the administration was allegedly pressuring Rick Bright, responsible for vaccine development as the head of the Department of Health and Human Services’ Biomedical Advanced Research and Development Authority (Barda), to get on the hydroxychloroquine train. According to Bright’s eventual whistle-blower report, the general counsel for HHS told Bright’s team that the White House wanted an Investigational New Drug protocol for chloroquine to accommodate a soon-to-come donation of millions of doses from Bayer. Bright managed to talk his bosses down to an emergency use authorization, a less full-throated support of the drug’s efficacy. “When I resisted efforts to promote and enable broad access to an unproven drug, chloroquine, to the American people without transparent information on the potential health risks, I was removed from Barda,” Bright told a subcommittee of the House Energy and Commerce Committee.
On March 27, the FDA announced an emergency use authorization for hydroxychloroquine and chloroquine to treat Covid-19, freeing up the drugs for use on sick patients. Prescriptions skyrocketed, mostly from physicians who’d never prescribed it before. Many people who volunteer for clinical trials do so out of community spirit; some also hope to get access to a potentially crucial drug—risking the chance that they might instead get randomized to the placebo group. Widespread availability of hydroxychloroquine meant nobody needed to be in a trial to get it. The authorization had the counterintuitive effect of undercutting the effort to find out if the drug was actually worth taking.
Back in Minneapolis, Boulware suddenly found he couldn’t enroll enough people to get the statistical power his protocol needed to give a definitive answer. The research was on outpatients, people who weren’t hospitalized, all over the country—they could volunteer from anywhere. And the emails just stopped coming. Boulware read all the same news reports as everyone else. He could understand why. “Half of the people think it’s an unethical trial because it clearly works,” he told me in April, “and the other half thinks it’s clearly dangerous and we shouldn’t do it.”
They had 1,200 people enrolled. They only needed 180 more. They were so close.
The president’s advocacy added another, hyper-partisan political layer of difficulty. Trump supporters began to see the use of hydroxychloroquine, like the avoidance of wearing masks, as a badge of political allegiance. Even gentle cautions about potential bad health outcomes from hydroxychloroquine came to signal disloyalty. Drug companies weren’t pushing for trials. (Sandoz, a drugmaker with a business in generic, off-patent drugs like hydroxychloroquine, tried to mount a trial but canceled it for lack of participation.) The government wasn’t pushing for one, as it had for remdesivir. All of that left Boulware’s team hanging. Even his volunteers were telling him how they felt. “By mid-April, people had formed an opinion,” he says. “Either it worked or it was dangerous, and our enrollment was minimal.”
I asked Boulware if that’s normal, that participants in a clinical trial might have an opinion about whether the drug they were testing worked or not.
“No,” he said, “it’s not normal, but I guess I’ve never been involved with a clinical trial that became political. I don’t think any clinical trial has ever been political while it was ongoing.”
The benefit of the doubt and goodwill toward others that clinical researchers depend upon in their volunteers was gone, thanks to the president. “What do you have to lose?” Trump said at a press briefing in April. “We don’t have time to go and say, ‘Gee, let’s take a couple of years and test it out. And let’s go and test with the test tubes and the laboratories.’” Days later, two former FDA commissioners went on record saying the emergency use authorization had been a terrible idea, because of the lack of efficacy data. But the president had no interest in slowing things down.
That kind of cavalier approach—hey, why not?—puts physicians in the position of balancing a chance of benefiting the patient in front of them against the certainty of not benefiting patients in the future. It’s a terrible choice. It also exists in a fog of privilege. Only people rich enough or with good enough insurance can afford, literally and metaphorically, to make a mistake. If the drug helps, they got it before anyone else could. If it does nothing, no matter. And if it does harm, well, they have access to medical care to save them. The whole concept seems like it gives individuals autonomy, but making a decision with insufficient information isn’t autonomy. It’s desperation, and it comes at the expense of everyone who gets sick later. This dangerous tactical individualism degrades both personal responsibility to community and overall scientific knowledge. Sick people become panicky nihilists, and no one ever learns anything.
Since the 1970s, a certain lineage of epidemiologists had been arguing that really massive randomized controlled trials could provide a scientific bulwark against that egotistical nihilism. When most drugs have only moderate-size benefits, you need thousands of people in the trial. When scientists and companies are motivated by social and commercial needs to get positive results, you need randomization to get good evidence. It’s the only way to change policy and treatments.
At least, that’s what an Oxford researcher named Martin Landray had come to think. A professor of medicine and epidemiology and acting head of the Big Data Institute at Oxford, Landray made his bones on large-scale cardiac trials; recently, he’d been working on policy, trying to simplify the regulations around those kinds of big studies. The Covid-19 pandemic gave him a chance to put the idea into action. Just a couple of weeks after Boulware put his hydroxychloroquine protocols together, Landray and Peter Horby, an expert in conducting trials during epidemics, built something bigger. Much bigger.
The Randomised Evaluation of Covid-19 Therapy Trial, also called Recovery, would split thousands of Covid-19 patients into groups testing various drugs as soon as they entered a hospital. Just about every other aspect of the UK’s Covid-19 response has been, in the local argot, a massive cock-up, but this thing they got right. Landray and Horby got approval to build patient consent for the study into hospital admission processes across the country. The National Health Service’s electronic medical records made it easy to track what happened to people with Covid, and the outcome they decided to measure for every drug on the roster was the simplest one: mortality. Did people, simply, die? “When you’re in a pandemic, just thrashing about is not helpful. One has to actually go back to the basic principles of randomized trials to determine which treatments work and which do not,” Landray says.
The first drugs they picked were already available, but no one was clear whether they worked. Dexamethasone, a steroid, was controversial because of the double threat of Covid-19. In early stages, the disease acts like a run-of-the-mill virus, damaging cells, especially in the lungs. But in the second stage of the disease, a person’s own immune system overreacts, causing widespread damage and sometimes death. Steroids are immunosuppressants that can calm that overreaction but also tamp down the good immune response. So it wasn’t clear whether a steroid would help the second phase more than it harmed the first.
“I mean, can you imagine being Trump’s doctor? Clearly Trump wants it, and he’s going to get it no matter what. It’s hard to say no to that.”
The other drug candidate was a combination therapy of HIV antivirals many hospitals were relying on—including Montefiore Medical Center.
At first Landray and Horby didn’t include hydroxychloroquine. They added it in April. “It was a choice a lot of people were interested in,” Landray says. “And if it wasn’t in the trial, a lot of people were going to use it anyway.” (Landray was aware of the “circus” in the US, but people elsewhere were advocating the drug too. “I’m not just talking about the president of the US,” he says. “He’s been a high-profile advocate of all sorts of things.”) All they needed was a couple thousand people taking hydroxychloroquine, and up to 4,000 who were not, and they could rule it in or out.
Back in Minnesota, it wasn’t until early May that Boulware’s team managed to eke out enough participants for statistical significance. They wrote up the results in three days, a dozen people sharing one Google Doc, and they sent two papers to The New England Journal of Medicine. Both showed negative results. Hydroxychloroquine didn’t ease symptoms any better than a control, and it didn’t prevent anyone from getting sick after exposure to an infected person. The papers weren’t perfect, but the data was clear: The drug didn’t work. Then, on the same day he submitted the papers to the NEJM, “I got an email from the White House asking about post-exposure prophylaxis,” Boulware says. “It was a memorable day.”
The public didn’t know it yet, but one of President Trump’s valets had tested positive for Covid-19. The White House staff knew Boulware had been working on post-exposure prophylaxis, and the president’s doctor wanted to see the trial results. “If it were normal times, I would say sure, that’s fine,” Boulware says. But NEJM follows something called the Ingelfinger rule, named for a preeminent early editor, that says if your data has been reported or submitted somewhere else, you can’t also publish it in NEJM. Boulware was worried that the White House might release the data and screw up his chances with the journal.
So Boulware demurred to the White House. He told the staffers that his team’s analysis was still ongoing. “I did say, ‘Based on the data we’re aware of, we don’t recommend this,’” he says. “I gave a recommendation based on my judgment.” But Boulware also told the White House doctor that it was safe to take, at least.
“I mean, can you imagine being Trump’s doctor? Clearly Trump wants it, and he’s going to get it no matter what,” Boulware says. “What he wanted to do and what he thought was the best judgment versus the president of the United States? It’s hard to say no to that.”
Several days later, the president announced at a press conference that he was indeed taking hydroxychloroquine. “The president has always said that he sees hydroxychloroquine as a very promising prophylactic, but that it should only be taken in consultation with your doctor,” Sarah Matthews, a White House spokesperson, told me in an email. “The president has personal confidence in it, as he has taken it himself as a prophylactic.”
A couple weeks after Boulware told the White House doctor that hydroxychloroquine was safe even if it didn’t work, the respected medical journal The Lancet published the results of a study erasing even that silver lining. It wasn’t a clinical trial. It was, on its face, an observational study reviewing outcomes from nearly 100,000 Covid-19 patients on six continents. As big data goes, that was pretty big. The authors said their data showed that the drugs caused a significant increase in potentially fatal heart problems, a risk that could outweigh any benefit. The impact of the paper was huge. Within a few days, the World Health Organization announced it was pausing the hydroxychloroquine arm of its major study. Regulatory agencies around the world started making noise about canceling more studies, revoking use authorizations.
Landray wasn’t convinced. “I was mildly irritated, disappointed, that people were taking the paper seriously, because it was an observational study,” Landray says. “The people who got the drug are different from the people who didn’t, in all sorts of ways you can’t measure or successfully disentangle.” But the tangle was real nevertheless. UK health regulators wanted to know what was going on; at their behest, Landray asked his data monitoring committee to take an unscheduled look at their findings so far—without letting him or any of the other researchers see it—for signs of clear benefits or harm.
In fact, the Lancet paper was sitting poorly with lots of people. In Thailand, a malaria researcher named James Watson read it on a Friday night, after he’d put his kids to bed. “My first thought was, this effect on cardiotoxicity seems too big to be real,” Watson says. He’s a senior scientist at the Mahidol Oxford Tropical Medicine Research Unit, and at the time he was working on pharmacology for a hydroxychloroquine study. To him, the statistics in the Lancet paper looked hinky. The paper didn’t even mention the most dangerous kind of cardiac arrhythmia that hydroxychloroquine can cause. “The most important data was missing,” Watson says.
The next day Watson’s boss got a phone call. Health regulators in the UK were suspending their study. The team was shocked. It seemed wrong. They had an emergency meeting—it was Saturday—to reverse engineer the paper. They thought it must have had methodological flaws. They were wrong, though. The actual explanation was much, much worse.
Over the course of the following week, Watson exchanged emails with The Lancet and with the paper’s lead author, an illustrious cardiologist at Brigham and Women’s Hospital in Boston named Mandeep Mehra. The data, it turned out, came from a company called Surgisphere, a slightly mysterious 13-year-old company with scant history of working with patient medical records. People started noticing some flaky stuff almost immediately: The data was aggregated not by country of origin but by continent. But a reporter at The Guardian noticed that the Australian data didn’t match that country’s Covid-19 stats. “We started thinking, maybe the data are rubbish,” Watson says. He wrote an open letter demanding clarification from the journal and authors, and hundreds of researchers signed it—including Boulware. “Everyone had read this paper, everyone had seen different difficult, weird parts of it,” Watson says.
The pro-hydroxychloroquine forces were just as activated. James Todaro, the guy who wrote that first white paper, wrote another one: “A Study Out of Thin Air,” in which he too laid out all the very real problems with the paper. It was a double-helix of irony. By now lots of researchers suspected the drug didn’t work, but they were criticizing a bad paper that said so; supporters of the drug’s use were touting the bad paper as evidence of unfair suppression of an effective medicine.
Mehra, through a spokesperson, declined to be interviewed or to answer emailed questions; he told The Scientist that he hadn’t been aware of any problems with Surgisphere’s data before publication and referred other questions to one of the other authors on the paper. That author has since been terminated from an adjunct position at the University of Utah, and the third author—Surgisphere’s founder, Sapan Desai—has also left his job at a Chicago hospital.
No one actually knows for sure what went wrong with any of the papers that used Surgisphere’s numbers, but it seems clear that the underlying data from Surgisphere doesn’t represent actual outcomes from actual patients taking actual hydroxychloroquine. Perhaps the journal moved too fast, failing to enforce standards of responsibility for data upon its authors. Everyone was operating in a high-velocity moment in which every new bit of Covid-19 information got picked up and picked over by the scientific establishment and the mainstream press. Desperation made them all vulnerable.
Landray’s data monitors didn’t find people suffering from heart problems, and his trial continued. “That, I think, was an important decision, because with a drug that was being so widely used, it’s really important to get the right answer,” he says. “Even at that point I thought, it’s possible this treatment might work. We don’t know.”
Then things accelerated. Over just a few days, Boulware’s first paper came out. The Recovery trial announced it was canceling its hydroxychloroquine arm, not because the drug was dangerous but because an analysis of the data showed that it did no good. The WHO, which had restarted its study after the Surgisphere mess, shortly thereafter re-canceled it for the same reason as Recovery. So did the NIH.
The Lancet retracted the Surgisphere paper—which had the confounding effect of making hydroxychloroquine seem good to its proponents, including the president of the United States. Matthews, the White House spokesperson, cited the retracted paper to me as an example of “misleading studies out there that were heavily touted by the media.” Yet, as a capper, the FDA revoked the emergency use authority for the drug. A few smaller studies are still ongoing, and technically physicians can still prescribe the drug off-label—but the tens of millions of doses in the stockpile are now a no-go for Covid-19. After all, it doesn’t work.
Hah, no, just kidding! Of course that wasn’t the end. The large clinical trials did manage to get hydroxychloroquine out of the running to be part of the standard of care for Covid-19. Some researchers still think the drug might have a small, as yet unproven effect if used early enough, or in a different amount. It’s possible, and it’s also possible no one will ever know.
That would be normal. Part of knowing how to know stuff is knowing what the edges are. All science is settled, until it isn’t.
In July two more big randomized trials hit that showed hydroxychloroquine having no effect. That didn’t stop White House economics adviser Peter Navarro from touting the drug on TV. The propaganda site Breitbart, which had been an early proponent, posted a video from a group calling itself America’s Frontline Doctors, which likewise praised hydroxychloroquine and described its demise as the result of an “orchestrated attack.” The president and his son both shared the video. So did Madonna. One of the main speakers in the video turned out to be a doctor with a storefront clinic that was also a church. It quickly emerged that she wrote a book about illness being the result of demonic impregnation, which it is not.
By the way, that line about an orchestrated attack came from the “investigative physician” of America’s Frontline Doctors—James Todaro.
The science infrastructure of the federal government might have been able to head off all this politicization and weirdness. A simple message of calm, plus the coordination of actual clinical trials, could have cleared away the confusion and ambiguity. But that didn’t happen. Such actionable information could have stood in the way of the, uh, nonscience infrastructure doing whatever it was they wanted—to prove that they were smarter than scientists, to show that there was a miracle cure, to sow political chaos. In the middle of a pandemic that killed more than a quarter-million Americans, that waste of time was a waste of human life.
As a coda to all this, a funny story: About 6,000 words ago I mentioned that some of the earliest evidence that hydroxychloroquine and chloroquine could help with the fight against Covid-19 came from in vitro trials—mix a little of the drug with some virus and some cells in a petri dish and see who wins.
Well, in late July a team of German researchers pointed out that early, seemingly successful tests of chloroquine used a cell line that’s derived from the kidneys of African green monkeys. SARS-CoV-2 affects lots of different organs, including the kidneys, but its primary target is the lungs. So the German researchers got a culture of lung cells and exposed them to the virus—and to both hydroxychloroquine and chloroquine. Neither drug did a bit of good. None. Bupkiss.
Boulware’s team had been working on one other trial since April. It was for “pre-exposure prophylaxis.” They gave the drug to health care workers before they were exposed to Covid-19, to see if it kept them healthy. When we talked about it, Boulware seemed to care a little less about what the outcome would be this time. He’d had enough. “If it doesn’t work, we’re going to be, like, that’s fine. We’re kind of burned out. Let’s just get it done, write it up, publish it, and move on, because we don’t like the political aspect of any of this,” he says. (The results came out in October; the drug didn’t work.)
A coda to the coda: In the early morning hours of October 2, Donald Trump announced that he had tested positive for Covid-19. Amid a fog of disinformation about his condition and treatment, his doctor released a list of the drugs they were giving him, including the antiviral remdesivir, still-experimental and unapproved monoclonal antibodies, and untested but potentially useful things like zinc and vitamin D. Hydroxychloroquine wasn’t on the list.
All Rights Reserved for Adam Rogers