By March 29, 2020, Covid-19 had afflicted more than one hundred thousand people and killed more than three thousand in the United States alone.1 On that same day, Dr. Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases and spokesman for the White House Coronavirus Task Force, predicted that the pandemic would kill two hundred thousand Americans.2 By this time, states had instituted lockdowns; major events were postponed or canceled, including the 2020 Summer Olympics; schools were closed or moved online; and businesses were shuttered.3 People prepared to shelter in place, sparking shortages of common goods such as toilet paper. Doctors sought to discover the best options to diagnose and treat patients infected with the virus. In the face of widespread shortages, hospitals scrambled to buy gowns, masks, gloves, and hand sanitizer.4
As a physician, I worried for my own safety and that of my patients. I spent almost every free moment reading about, thinking about, and communicating with other medical professionals about treating patients infected with the virus. I joined Facebook groups in which physicians from North and South America, Europe, Asia, and Australia posted new information day and night. I spoke with colleagues in the first two hot spots in America, Washington State and New York City, to learn from their experiences. I read old publications on SARS and MERS and new ones on Covid-19 from China and Italy.
Early in the pandemic, physicians tried many treatments. One such treatment, a now infamous and controversial drug, was known to have some antiviral properties from decades-old studies.5 This drug, hydroxychloroquine (HCQ), was and still is used by physicians worldwide to treat Covid-19. On March 20, French researchers and physicians published data suggesting that HCQ could reduce the virus’s ability to replicate.6 Graphs from this paper immediately circulated among physicians, bolstering excitement for the drug, which led to increased use. . . .
The following day, on March 21, President Trump tweeted, “HYDROXYCHLOROQUINE & AZITHROMYCIN, taken together, have a real chance to be one of the biggest game changers in the history of medicine.”7 A week later, on March 28, the FDA issued an emergency use authorization (EUA) that allowed medical professionals to tap into the Strategic National Stockpile (SNS) of HCQ when prescribing it for the treatment of Covid-19. On March 29, the U.S. Department of Health and Human Services (HHS) accepted a donation of thirty million doses of HCQ, which were added to the stockpile.8 States, hospitals, and individual physicians also bought the medication in bulk.9 Even while excitement for the treatment spread, Dr. Fauci already was cautioning that no strong evidence indicated that it was effective for treating Covid-19, and FDA commissioner Stephen Hahn said it was possible the drug could do more harm than good.10
On April 22, the Veterans Health Administration released a study showing that HCQ did not benefit hospitalized patients who were treated with it.11 Then, on April 24, the FDA issued a warning suggesting that HCQ, particularly in combination with azithromycin, may put patients in danger, especially if used outside the hospital in an unmonitored setting.12 On May 7, the New England Journal of Medicine (NEJM) released another study that did not show a clear benefit of HCQ for the patients studied.13
On May 18, despite the FDA’s safety warnings and data mounting against HCQ for hospitalized patients, the New York Times reported that President Trump was taking HCQ “as a preventive measure.”14
On May 22, The Lancet published a study suggesting that using HCQ is “associated with an increased risk of de-novo ventricular arrhythmia during hospitalisation.”15 Ventricular arrhythmias are abnormal heart rhythms that, if not resolved, can result in death. Given this reported risk, the World Health Organization (WHO) ceased its international HCQ trial. However, the trial was resumed in June after it was discovered that the underlying data from The Lancet study likely had been fabricated by a dubious firm named Surgisphere.16
Less than two weeks later, on June 15, the FDA revoked the EUA that had allowed physicians to prescribe HCQ for Covid-19 from the SNS.17 After this and the FDA’s warnings, many hospitals ceased the routine prescription of HCQ, mostly limiting the drug to patients enrolled in clinical trials.
In addition, at least one state attempted to ban the use of HCQ for Covid-19. On July 29, the State of Ohio Board of Pharmacy temporarily prohibited pharmacists from selling or dispensing HCQ for Covid-19, a ban that lasted only two days and was overruled by the governor after a public backlash.18 Many other state pharmacy boards issued statements and policies restricting, to varying degrees, the use of HCQ for Covid-19.19 Researchers and medical professional societies recommended against using HCQ, and public messaging from these same groups warned patients and physicians of the possible dangers of arrhythmia and sudden death.20 By July, these combined actions made it very difficult for patients with Covid-19 to receive or fill a prescription for HCQ.
Nevertheless, a vocal minority continued to advocate the widespread use of HCQ for treating and/or preventing Covid-19. Some politicians were persistent in their advocacy; on Twitter, President Trump shared a video of physicians who called themselves “America’s Frontline Doctors,” extolling the virtues of the drug.21
Some who used social media to advocate the use of HCQ had their content removed by Twitter, YouTube, and Facebook on the grounds that it violated the misinformation policies of those platforms.22 In a statement to BBC News, Twitter said, “Tweets with the video are in violation of our Covid-19 misinformation policy. We are taking action in line with our policy here.” Facebook told the BBC, “We’ve removed this video for sharing false information about cures and treatments for Covid-19.” And YouTube told the BBC, “We have removed the video for violating our Covid-19 misinformation policies.”23
As a result of this social media suppression, the involvement of politicians, and the Surgisphere research scandal, many became suspicious that HCQ had fallen out of favor in the medical community for political rather than medical reasons.
With this inconsistent and controversial public narrative, the drug remained at the forefront of many people’s minds. Understandably, when some of these people were diagnosed with Covid-19, they sought prescriptions for HCQ. Others who weren’t sick, but who understandably were upset about the prolonged economic and social impacts of government-enforced lockdowns, became frustrated with the idea that a potentially efficacious drug was being withheld from the public.
On July 23, 2020, Newsweek published an article titled “The Key to Defeating Covid-19 Already Exists. We Need to Start Using It,” by Dr. Harvey Risch, a professor of epidemiology at the Yale School of Public Health and Yale School of Medicine.24 Therein, Dr. Risch said of HCQ:
I am fighting for a treatment that the data fully support but which, for reasons having nothing to do with a correct understanding of the science, has been pushed to the sidelines. As a result, tens of thousands of patients with Covid-19 are dying unnecessarily. Fortunately, the situation can be reversed easily and quickly.25
The article reignited public discussion of the drug. My patients, friends, and family bombarded me with questions, asking if Dr. Risch was right. So, I read his Newsweek article and thought it was the most convincing opinion I’d seen in the HCQ debate. He made points and distinctions that others ignored. I thought, if he’s correct, we may have a widely available and effective treatment for Covid-19.26
In order to fully understand Dr. Risch’s claims, I reviewed all of the evidence he cited in his two articles for the American Journal of Epidemiology (AJE);27 I reread his article in Newsweek; I watched his appearances on Ingraham,28 Hannity,29 War Room,30 and the John Berman Show;31 and I corresponded with him via e-mail in search of answers to some of my questions.32 I also reviewed the arguments of other prominent HCQ advocates before choosing to focus on Dr. Risch’s argument; his is the clearest and most thorough evaluation of the evidence in favor of HCQ.
In essence, he argues that HCQ:
Further, he holds (6) that these drugs, even when given in combination, almost always are safe. (His protocol stipulates the outpatient setting because, typically, a patient admitted to the hospital for Covid-19 treatment is no longer in the early stages of the disease.) On this basis, Dr. Risch argues that using his HCQ protocol immediately and widely to treat Covid-19 patients would save tens of thousands of lives.
In order to evaluate this claim, I reviewed every citation in Dr. Risch’s publications, and I continued to review new evidence until the publication of this article. After careful review of the studies that appear to show positive results for HCQ, I do not think the current evidence supports this conclusion. All of the studies that I have reviewed suffer from significant methodological flaws that call the apparent benefit of HCQ into question.33 Of course, we’re learning more about Covid-19 every day, and it may turn out that HCQ is a useful treatment. However, I don’t believe today’s best evidence supports this conclusion.
Upon review of Dr. Risch’s citations, I found only one study that directly addresses his hypothesis. The other citations include patients who do not meet his full criteria. Several others are not studies at all but local news stories34 or personal communications that have no published protocol, methodology, or data. The citations that are studies all share at least one major methodological flaw that makes causal inference impossible: In each study, researchers did not properly control or balance relevant variables.
Even the research that best supports using an HCQ-based protocol, a study from Brazil,35 contains this problem. This study is the strongest because, of all of the studies that appear to demonstrate a benefit for HCQ, it’s the only one that follows Dr. Risch’s protocol and compares the treatment group with a control group. This study demonstrated that 1.9 percent of those treated with HCQ and azithromycin were hospitalized, versus 5.4 percent of those untreated.
At first this sounds promising for HCQ, but this study has major flaws. In order to pinpoint the factor(s) that led to different outcomes for these groups and thereby demonstrate the efficacy of the treatment protocol, the study would have to control or at least balance other relevant variables. But this study did neither. Patients with flu-like symptoms were offered HCQ and azithromycin, and those who refused the drugs (and so were not treated with them) constitute the study’s nontreatment group (subsequently referred to as the “refusal group”). No documentation indicates why this group refused treatment, and it’s possible that many refused the HCQ treatment because they had symptoms more indicative of other illnesses.36
Consider that a patient who was vomiting and experiencing abdominal pain could have been included in the study on the grounds that he had flu-like symptoms. But he may have refused the HCQ protocol on the grounds that his doctor thought his symptoms more indicative of an infection of the appendix, gallbladder, kidney, or colon. Note, too, that these conditions commonly require hospitalization.
This leads us to what is, far and away, the biggest problem with this study: The patients weren’t tested for Covid-19. Obviously, we can’t learn anything about the efficacy of a treatment for Covid-19 if we don’t know if the patients being studied even have the disease. It’s possible that the study is comparing patients with Covid-19 in one group to patients with completely different medical problems in the other group or, more likely, that both groups include a mixture. Consider, for instance, that azithromycin—by itself—is a common and efficacious treatment for several types of pneumonia and other infections.37 It’s possible that people in the treatment group had one of these infections, not Covid-19, and that the azithromycin cured them. Similarly, it’s possible that patients in the refusal group also had one of these infections and that, in turning down the HCQ+azithromycin, they prolonged a condition that would have improved with azithromycin. It would take only five such patients in each group to yield equal hospitalization rates for both groups.38 And, as noted above, it’s likewise possible that a significant number of those in the refusal group did not have Covid-19, in which case, the fact that they subsequently were hospitalized isn’t relevant to an evaluation of an HCQ-based protocol for Covid-19.
On September 15, I e-mailed one of the authors of this study for more information but have not yet received a response. I also asked Dr. Risch for his thoughts on the shortcomings of this study, to which he replied:
In this study, the treatment group was older and had higher frequencies of diabetes, stroke and hypertension, and greater progression of disease with higher frequencies of fever, cough, anosmia [loss of smell], diarrhea, headache, myalgia [muscle pain] and dyspnea [difficulty breathing]. The confounding e in this study is thus demonstrably against a beneficial effect of the medication, yet the study shows just such an effect, significantly so.39
In other words, Dr. Risch is saying that it’s remarkable that the treatment group had fewer hospitalizations given that it was composed of more symptomatic and higher-risk patients compared to the refusal group.
If Dr. Risch is right, and if all or most of the participants had Covid-19, then I believe this is the strongest evidence in favor of Dr. Risch’s HCQ protocol. But even if we accept this assumption, the study still demonstrates only a decrease in the rate of hospitalization (1.9 percent for those in the treatment group and 5.4 percent in the refusal group)—not a decrease in death rate. Only one patient from each group died out of a total of 636 patients, and neither death was due to Covid-19. So, even if this study’s results are reliable, it does not support the claim that we can save one hundred thousand lives with an HCQ-based protocol. In order to support that claim, the study would need to demonstrate that such a protocol actually saves the lives of those diagnosed with Covid-19. Instead, it demonstrates, at best, that the mortality rate for the patients involved—some of whom likely had Covid-19 and others who likely did not—was exceedingly low, regardless of treatment. Speculation aside, the study demonstrates nothing conclusive about HCQ for the treatment of Covid-19, as its participants were never confirmed to have the disease.
Other studies suffer from similar problems. For example, in one study that appears to show a benefit for HCQ, the authors compare a treatment group that includes only ten patients over age sixty-five with a nontreatment group that includes 154 patients over age sixty-five. The nontreatment group also includes those unable to take HCQ because of the likelihood of complications due to other illnesses or conflicting medications. Further, the nontreatment group includes more than double the number of patients with diabetes (402 vs. 171), hypertension (342 vs. 131), and “heart diseases” (166 vs. 82).40 The specific medical problems that usually preclude the use of HCQ—problems with the heart, blood vessels, and lungs—make any patient more susceptible to hospitalization and death from Covid-19, or any other heart or lung infection. This means that nontreatment group patients were at higher risk for hospitalization and death to begin with. The fact that they didn’t receive HCQ was a result of their poor prognosis, not the cause of their hospitalization or demise.41
In an attempt to find the strongest data, I asked Dr. Risch to indicate what he thought were the top three studies that support his position, and in response he said, “I don’t subscribe to the pick-the-best-studies-and reason-from-them analysis. One has to form a gestalt over all of the evidence and ask whether all of the evidence could have arisen by chance (or artificial contrivance if manipulation were involved) if the association really didn’t exist.”
I agree that we need to examine the totality of the evidence. But, we also need to evaluate the quality of the evidence. If a study has unclear or problematic methodology that could affect its outcome, it cannot be considered strong evidence for or against a treatment. Thus far, this is the case for all of the studies I’ve reviewed that appear to support the use of HCQ for treating Covid-19. And, in most of these studies, the authors themselves identify shortcomings and conclude that further research is needed.
If we are to consider the totality of the evidence, we also must analyze randomized controlled trials (RCTs) that have tested HCQ for the treatment of Covid-19. To date, at least five RCTs have not found a benefit of HCQ in a variety of circumstances, though none test Dr. Risch’s treatment criteria in full.42
What is the big deal about RCTs, and why should we consider them? An RCT is a trial in which patients are enrolled and then randomly assigned to one of at least two groups—usually a treatment group and a control group. Sometimes a study has multiple treatment groups. The purpose of randomization is that it eliminates many biases and balances variables that might otherwise affect results. As Dr. Risch’s Yale colleague Perry Wilson states, “If you randomize, you balance not only measured differences, but even unmeasured differences between groups.”43
RCTs are not necessary to prove the efficacy or safety of simple treatments in which cause and effect can be plainly observed. Consider applying pressure to a moderately bleeding wound, for instance. Anyone can see that applying pressure to such wounds almost always reduces or stops the bleeding. This is evident by direct observation, so no RCT is needed to learn whether this technique works. But, when the effect of a treatment becomes smaller in magnitude and delayed in time, when its mechanism is invisible to the naked eye, or when a plethora of variables are involved, it’s much harder to tell which causes are responsible for a particular outcome. RCTs are one tool that can help isolate causes and measure their effects amid numerous variables.
Without RCTs, it can be difficult or impossible to connect certain causes and effects with any degree of certainty. For instance, in an HCQ study done at the Henry Ford Health System, treatment groups were not randomly selected. Rather, patients, in consultation with clinicians, chose whether to take HCQ. Those choices were purposeful and likely based on a variety of factors (e.g., baseline health and quality of life, coexisting medical conditions, quality of care, etc.). As a result of these factors, these two groups were qualitatively different at the outset. Many subsequent treatment decisions were likely affected by these differences. Even if HCQ was responsible for the difference in outcomes, it’s impossible to be certain given the numerous confounding variables—the same variables that RCTs are designed to minimize.
For this reason, many physicians focus on RCTs when reviewing medical evidence for treatments such as HCQ.44 In his first AJE article, Dr. Risch agrees that RCTs are the highest quality evidence, and in his interview with John Berman, Dr. Risch said,
Randomized controlled trials, in the real world, are not gold standards. That is a theoretical idea that does not play out in the real world. For example, in sepsis studies, nobody uses randomized controlled trials for medications to treat sepsis. That’s standard, and the whole field knows that. Dr. Frieden [director of the CDC] explained that all real-world evidence goes into understanding the benefits and harms of medications, and that is the standard that scientists use. Relying on a so-called theoretical randomized controlled trial is a red herring because that is not the way the FDA works, it is not the way scientists work, it is not the way anybody works in the real world.45
I agree that RCTs are not immune to methodological flaws. Their results also can be misleading or misapplied. And physicians do make many reasonable decisions without the benefit of RCTs. But it is important to note that researchers have conducted many RCTs to evaluate medications for the treatment of sepsis.46 The FDA does rely on RCTs, as do most physicians, when evaluating new treatments. And, professional societies create guidelines based on RCTs, which are used widely in every specialty of medicine to guide diagnostic and treatment decisions. In short, when methodologically sound and conducted in the right patient population, RCTs are one of the best and most widely used tools that medical professionals have for evaluating the efficacy and safety of medical treatments.
Although Dr. Risch states in his first AJE article that RCTs are the highest form of medical evidence, he argues that we can’t wait for the results of further RCTs to take action. In essence, his argument is: RCTs are the best form of evidence, but they are not the only form of evidence. The preponderance of the existing evidence supports the use of HCQ for Covid-19, and we can’t afford to wait for the perfect RCT, which has not yet been done. Therefore, on the basis of the available evidence, we should advocate the immediate and widespread use of HCQ.
I agree with Dr. Risch that RCTs often are imperfect, are not the only valid form of evidence, and that we must weigh all available evidence—RCT or not. And he is right to point out that the RCTs studying HCQ have flaws. For instance, one RCT has a problem similar to that in the Brazil study discussed above; only a minority of its subjects were tested for Covid-19. However, in that RCT, medical professionals confirmed that each subject was exposed to Covid-19 and made their diagnoses based on specific clinical criteria. This is a superior methodology for ensuring inclusion of only patients with Covid-19 than was used in the observational study from Brazil.47 Even if the RCTs conducted thus far are flawed and thus have not demonstrated conclusively that an HCQ-based protocol doesn’t work for treating Covid-19, the burden of proof remains with those who claim that it does work.
In support of HCQ, some have pointed to international use patterns of the drug, arguing that countries with widespread use have fewer cases and deaths. One website has a map showing this, but I have been unable to track down its source data—or any other data that supports it.48 Regardless, given the myriad confounding variables, such patterns simply can’t tell us anything about the effectiveness of HCQ. These variables include variations in public health policy (e.g., masking, social distancing, lockdowns, etc.), genetics of different populations, climate, differences in sun exposure and nutrition, social practices, living conditions, the use of other drugs, the state of the health system, hospital protocols, baseline population-wide health and hygiene, and many others. The variables involved in such an international comparison make it impossible to isolate HCQ as the decisive causal factor in Covid-19 outcomes.
In total, the data used to support an HCQ-based protocol for treating Covid-19 suffers from a multitude of problems. I hope that Dr. Risch and others ultimately are right about the efficacy of HCQ, but I cannot reach the same conclusion given the current state of the evidence.
I do agree with Dr. Risch and other advocates of HCQ on two crucial points. First, HCQ likely is safer than it has been made out to be. It’s possible that HCQ and azithromycin, especially if given together to patients with certain conditions, may cause abnormal heart rhythms and, in some cases, sudden death. But the evidence indicates that such adverse effects are extremely rare, especially if given to patients who are not severely ill in the first place.49 Second, I agree that the FDA has no legitimate basis, medical, scientific, or otherwise, to restrict the use of HCQ.50
What should the FDA, state pharmacy boards, or any other government or regulatory agency do in regard to HCQ? They should create and enforce laws that protect individual rights—and nothing more. As long as manufacturers and salespeople do not violate people’s rights—for instance, by lying about the safety, efficacy, or some other aspect of the drug—government has no legitimate legal or moral grounds for restricting its sale or use.
In summary, HCQ initially seemed like a promising treatment for Covid-19. Dr. Risch, its most credible and strongest advocate, made a compelling argument in favor of its use amid political conspiracy theories and research scandals. However, after a thorough review, I do not think the current evidence conclusively demonstrates the efficacy of HCQ for the treatment of Covid-19. But whether or not HCQ is efficacious, the government must protect the rights of pharmacists, doctors, and patients to use their own judgment and engage voluntarily as they see fit.
if used with azithromycin or doxycycline and/or zinc,
given “early” to
symptomatic
“high-risk” Covid-19 patients (over the age of sixty and/or with comorbid conditions such as diabetes and hypertension)
in the outpatient setting, could save 75,000–100,000 lives.