What does it mean that Oracle is partnering with the Trump administration to study unproven COVID-19 drugs?

By Lisa Larrimore Ouellette, Nicholson Price, Rachel Sachs, and Jacob Sherkow

One of the dizzying stream of innovation and health law stories to emerge last week is Oracle’s partnership with the White House to study unproven pharmaceuticals for treating COVID-19. We decided to unpack this story for ourselves and then to collectively share our thoughts in a short explainer.

What are the drugs being studied?

The initial stories about Oracle’s platform mention its use for two older drugs approved to treat malaria—chloroquine and hydroxychloroquine—that are now being tested to treat COVID-19. Both drugs are quite old: chloroquine was first approved by the FDA in 1949 and sold, until recently, under the brand name Aralen. Hydroxychloroquine, which is also used to treat lupus and rheumatoid arthritis, is sold under the brand Plaquenil and was first approved in 1955.

The impetus behind studying these two drugs stems from in vitro studies following the 2005 SARS-CoV-1 outbreak. Those studies suggested the drugs could inhibit some types of coronaviruses from both entering cells and replicating after infection—potentially serving as a preventative and a treatment. But the studies were small, in cell culture rather than living animals, and not conducted against the virus that causes COVID-19, SARS-CoV-2. Some early work with the drugs against SARS-CoV-2 may be promising but it, too, has been done in a test tube rather than an animal model.

Importantly, though, chloroquine and hydroxychloroquine are not risk free. To the contrary, both have some significant side effects that may be especially concerning for COVID-19 patients. These include an increased risk of blindness and heart failure, the latter of which seems to be exacerbated by and is the primary cause of death in some COVID-19 patients. NPR recently reported that one COVID-19 patient on experimental chloroquine has died, seemingly as a result. It’s possible that chloroquine and hydroxychloroquine could make COVID-19 worse, not better.

More drugs besides chloroquine and hydroxychloroquine are being tested for COVID-19 treatment, and it is unclear which of these the Oracle platform might track. Drugs that have already received FDA approval for other indications include antiviral HIV drugs, interferons that activate the body’s immune response, and anti-arthritis drugs to treat lung inflammation. Researchers are also testing new drugs that have not yet received FDA approval; for example, the antiviral remdesivir, which is being scrutinized in numerous clinical trials, made headlines last week after Gilead asked for and received (and then withdrew) FDA designation of the medicament as an “orphan drug.”

When can doctors prescribe unproven treatments?

The FDA approves drugs for a certain set of uses (known as indications), but once a drug is approved, the FDA doesn’t limit how doctors use those drugs to treat patients. Instead, doctors can use drugs for purposes outside what the FDA has approved—so-called “off-label” uses. Many drugs are used off-label, and for some drugs, the drug is used more frequently off-label than on-label. Some of the easiest off-label uses to understand are for different populations; clinical studies in children are relatively rare, for instance, so many drugs aren’t approved for use in children. But doctors want to treat children, and so often use drugs off-label in children for the same condition. Other times, off-label use is for a different condition; a drug that treats one cancer might be used off-label to treat another. While some off-label use is supported by substantial scientific evidence, a disturbingly large fraction is supported by no or little good evidence.

There are some potential limits on off-label use. Insurers might not pay for an unapproved use, for example. Off-label use might also incur a higher risk of tort liability, and state medical or pharmacy boards may disapprove of a particular off-label use, making doctors wary of prescribing the drug off-label. For COVID-19, some state pharmacy boards have begun limiting the off-label use of hydroxychloroquine to prevent shortages for patients who need the product for its approved uses for lupus and rheumatoid arthritis. (In fact, to address shortages like this, FDA has now approved compounding of hydroxychloroquine, an alternate pathway to standard drug manufacturing where the final formulation steps are done by pharmacies or others rather than drug manufacturers.)

In some cases, drugs still under investigation can be used to treat patients under programs known as “compassionate use” or “expanded access.” The FDA’s expanded access program has eligibility requirements that include the lack of other available treatments, a serious disease, and an inability to get the same drug from another clinical trial or program. Remdesivir has, in fact. been used to treat some patients under this pathway, even though it is not yet approved by the FDA for any use.

In 2018 the federal government also passed Right to Try legislation, which lets drug manufacturers provide drugs to patients if the drug has passed the earliest phase of clinical testing and is in active development, without first asking for FDA permission. This pathway has been roundly criticized, but may not make much difference for COVID-19 given that it has been used quite rarely so far.

What kind of data might Oracle’s platform collect?

According to last week’s reports, Oracle’s new website and mobile app—which it plans to donate to the government—will be used to collect information about COVID-19 treatments outside the context of clinical trials. It is unclear whether this platform will in fact be developed; on Friday, Politico reported that “[t]he White House is having second thoughts about working with Oracle on the project after a New York Times article this week raised questions about the partnership.” But if the project (or another one like it) goes forward, how could this kind of data be used to help address the pandemic?

Our understanding is that Oracle’s platform would help facilitate observational studies, or investigations of data that are not gathered in a controlled experiment. Researchers could examine, for example, whether COVID-19 patients treated, off-label, with hydroxychloroquine have better outcomes than patients who do not receive it. Of course, as we explain below, interpreting these results is difficult—perhaps patients who receive hydroxychloroquine happen to be healthier (or sicker) than those who don’t—which is why drugs are typically tested in controlled clinical trials. But an observational study can be more informative than no study at all.

Observational studies are one kind of “real-world evidence” that FDA uses to inform regulatory decisions. In the 21st Century Cures Act, passed in December 2016, Congress directed the FDA to “evaluate and issue guidance on the use of evidence from sources other than clinical trials to support approval of a drug for a new indication”—i.e., a drug like hydroxychloroquine that could already be legally used off-label. In December 2018, the FDA published a framework for making these decisions, which notes that “observational studies may provide credible evidence” but cautions that there are “examples when effects identified in observational studies could not be reproduced in randomized trials.”

On Thursday, New York state was reported to be moving to distribute tens of thousands of doses of hydroxychloroquine and chloroquine (combined with the antibiotic azithromycin) to COVID-19 patients. Outcomes will be collected electronically, though it is unclear whether Oracle is involved in this collection. This work could inform other doctors who are debating whether to use these off-label treatments, and it could potentially support FDA approval for the COVID-19 indication under the 21st Century Cures Act framework.

How will reimbursement for this trial work?

The Washington Post reports that the administration is “exploring whether it will offer bonus payments to doctors who use the technology.” It is hard to determine exactly what the administration means by this, but in its most innocuous form, it might simply amount to committing to provide physicians reimbursement for prescribing these products for COVID-19. There are at least two pathways the federal government might use to do this, although both have limitations.

First, the New Technology Add-On Payment (NTAP) program administered by the Centers for Medicare & Medicaid Services (CMS) permits the agency to provide additional reimbursement for the purpose of encouraging the adoption and dissemination of new technologies in medical practice. However, the NTAP system is limited not only to the Medicare program but also even further to inpatient care, meaning that these payments would not be available to doctors prescribing these products to patients who are not eligible for Medicare or who are being treated on an outpatient basis.

Second, CMS’s Coverage with Evidence Development (CED) program permits the agency to provide reimbursement for a particular technology or service even if a clinical benefit thereof has not yet been established. In some ways, CED is much broader than NTAP, as it is not limited to the inpatient context. However, it is limited not only to Medicare, but also to situations in which clinical data are being collected to determine whether such a clinical benefit does exist—which may be the case under the Oracle program.

In either case, it is not clear how the administration would seek to offer bonus payments to physicians treating patients who are not Medicare beneficiaries. Private insurers could choose to cover these products (and indeed, some already have), but would not clearly be required to. Further, to the extent that the administration intends to do something other than provide reimbursement for these products, it is difficult to determine where true bonus payments might come from, as a statutory matter.

Shouldn’t these drugs be tested in randomized controlled trials?

Yes! Randomized controlled trials (RCTs) are the best way to find out whether a drug actually works in patients. It’s notoriously hard to know if drugs work; patients are different in many different ways, including the presence of other underlying medical conditions, and understanding whether a drug is, in fact, effective turns on who chooses to take a drug, who has access to it, what other medical treatments they might receive, and how well they adhere to the treatment regimen (among other things). RCTs are the gold standard for ironing out those sources of variation. They’re especially important when drugs have potentially significant side effects, as hydroxychloroquine and chloroquine do.

To be sure, RCTs of hydroxychloroquine and chloroquine are currently being conducted, though the results have been ambiguous so far. One study in France showed some promising results but has been criticized for its design. Another in China didn’t show statistically significant results. Both studies published so far have also been small (42 and 30 patients, respectively). Much larger studies coordinated by the WHO are underway for multiple potential treatments.

Is Oracle’s online platform a valuable complement to these trials or a waste of time?

As explained above, observational studies can be valuable complements to RCTs. With COVID-19 fatalities growing exponentially, doctors have to make treatment decisions before controlled studies are completed—and many may rationally decide that a treatment with weak evidentiary support is the best available option. Aggregating data from these individual decisions may provide some real-time evidence of which experimental drugs currently appear most effective.

But observational studies are not a substitute for rigorous RCTs, and we think there is a real risk that policymakers and the public will misinterpret their significance. Hype about hydroxychloroquine already far outpaces evidence of its effectiveness. For example, President Trump’s personal lawyer Rudy Giuliani tweeted that hydroxychloroquine “has been shown to have a 100% effective rate treating COVID-19,” and a conservative group took out Facebook ads and sent text messages asking supporters to join a petition to “CUT RED TAPE” around use of the drug. On Sunday, the FDA issued an “emergency use authorization” for hydroxychloroquine and chloroquine, permitting physicians to make them available for COVID-19 when a clinical trial is not an option for patients—a decision which was publicly criticized by a former FDA Chief Scientist, who decried the “total lack of scientific evidence” in support of the drugs’ efficacy.

In addition to creating a risk of policymaking based on misinformation, observational studies may decrease the urgency for RCTs. In a cautionary plea last week for enrolling far more COVID-19 patients in RCTs with adaptive designs, Andre Kalil worried that off-label and compassionate use of drugs “could discourage patients and clinicians from participating in RCTs, hampering any knowledge that could be gained about the effects of the drug being tested.” We think that this is a serious concern, and one that journalists covering stories about the Oracle platform and other efforts to administer observational studies should be asking about. In theory, tech companies like Oracle might even be able to help with administering and collecting data for RCTs. There is a place for big tech in this picture, and it may be a crucial one. But we shouldn’t let the excitement of collecting lots of observational data distract us from the reality that we still need a core of high-quality RCT data to know what works and what doesn’t.

Does Gilead's (withdrawn) orphan designation request for a potential coronavirus treatment deserve your outrage?

Many commentators were outraged by the FDA's announcement on Monday that Gilead received orphan drug designation for using the drug remdesivir to treat COVID-19. The backlash led to a quick about-face by Gilead, which announced today that it is asking the FDA to rescind the orphan designation. For those trying to understand what happened here and the underlying policy questions, here's a quick explainer:

How could the Orphan Drug Act possibly apply to COVID-19?

Under 21 U.S.C. § 360bb(a)(2), a pharmaceutical company can request orphan designation for a drug that either (A) treats a disease that "affects less than 200,000 persons in the United States" at the time of the request or (B) "for which there is no reasonable expectation that the cost of developing and making available in the United States a drug for such disease or condition will be recovered from sales in the United States of such drug." An ArsTechnica explainer suggests that remdesivir received orphan designation under option (B), but this email from the FDA indicates that it was option (A).

The designation seems correct based on the plain language of the relevant statute and regulations: As of Monday, there were 44,183 cases diagnosed in the United States (and even fewer at the time of Gilead's request), and the Orphan Drug Act regulations indicate that orphan designation "will not be revoked on the ground that the prevalence of the disease . . . becomes more than 200,000 persons." But given the CDC's low-end estimates of 2 million Americans eventually requiring hospitalization, commentators have noted that this feels like a loophole that gets around the purpose of the Orphan Drug Act.

What benefits would Gilead have received from an orphan designation?

The main effect would have been a tax credit for 25% of Gilead's expenses for the clinical trials it is running to figure out whether remdesivir is actually effective for treating COVID-19. (The tax credit was 50% when the Orphan Drug Act became effective in 1983, but was reduced to 25% by the December 2017 tax reform.)

Gilead would also have received a 7-year market exclusivity period if remdesivir is approved, but this would have had little practical effect because (1) it would already receive a 5-year data exclusivity period for approval of a new chemical entity (because remdesivir has not yet been approved for any use) and (2) Gilead has later-expiring patents, including a patent on the remdesivir compound that expires no earlier than 2035. In theory those patents could be invalidated, but patent litigation is a time-consuming process. So it does not seem likely that the 7-year orphan exclusivity would meaningfully enhance Gilead's pricing power during the peak of the COVID-19 pandemic.

Jamie Love of Knowledge Ecology International has argued that the government could use 28 U.S.C. § 1498 to overcome the patents and buy generic versions of remdesivir (an approach generally advocated by Amy Kapczynski and some of her students at Yale), and that there is no equivalent for regulatory exclusivity. But (1) 21 U.S.C. § 360cc allows generic approval of an orphan drug during the 7-year exclusivity if the sponsor "cannot ensure the availability of sufficient quantities of the drug to meet the needs of persons with the disease" (though invoking either this or § 1498(a) seems unlikely politically) and (2) as noted above, if remdesivir is approved, Gilead will still receive a 5-year data exclusivity period.

Should we celebrate that Gilead will no longer get this tax credit?

Not necessarily. It's true—as noted by commentators such as Laura Pedraza-Fariña—that the Orphan Drug Act seems like an odd fit for a COVID-19 drug, and that little about the orphan designation process requires an assessment of whether additional incentives are needed for developing a drug. And perhaps, as Mark Lemley suggests, Gilead will "go full steam ahead with trying to cure COVID-19" without an orphan designation. But "full steam" may not be a corner solution—there's usually a fuller steam.

There are of course a lot of uncertainties here, but the error costs seem very asymmetric: it is much worse to set incentives for addressing this pandemic too low than too high. The social cost of delaying access to a treatment—even by weeks or days—is huge; the cost of awarding an unnecessary tax credit to Gilead is just a transfer. There should probably be much more public funding for COVID-19-related R&D than there is, so allowing companies to be reimbursed for some of their clinical trial expenses may be the right move from a policy perspective. Giving Gilead an unnecessary transfer isn't ideal, but it seems pretty low on the list of things to be outraged about these days.

Right now, one of the first-order public policy concerns should be getting COVID-19 treatments and preventatives that are demonstrated to be safe and effective. Once those products are approved and marketed, widespread low- or zero-cost access will of course be important, particularly given the positive externalities associated with use. But as I have explained with Daniel Hemel, both in general and in the context of COVID-19 vaccines, innovation and affordability are not either/or. And innovation is the problem we should focus on first.

3/27/20 Update: Jamie Love thinks I am underestimating the importance of the extra exclusivity, and he could be right. The 5-year data exclusivity period for NCEs would prevent a generic manufacturer from obtaining approval based on Gilead's clinical trial data through the Hatch–Waxman process, but another firm could obtain a approval if it has enough clinical trial data to support its own application. But even if it is true that the more robust market exclusivity for orphan drugs would have added to Gilead's profits, that's not obviously a bad thing. There are so many anti-profiting-from-coronavirus stories these days, but I think we want COVID-19 to be the most profitable thing biomedical firms could be working on now. We don't yet have effective treatments, and patients desperately need them.

Challenging what we think we know about "market failures" and "innovation"

I really enjoyed the final version of Brett Frischmann and Mark McKenna's article, "Comparative Analysis of Innovation Failures and Institutions in Context." The article was published in Houston Law Review in 2019. But I initially encountered it when the authors presented an early draft at the 2012 Yale Law School Information Society Project's "Innovation Beyond IP Conference," conceived and brought together by Amy Kapczynski and Written Description's Lisa Ouellette. The conference explored mechanisms and institutions besides federal intellectual property rights (IP) that government uses, or could use, in order to achieve some of IP's stated goals. Examples explored include research grants, prizes, and tax credits, among countless others.

Frischmann and McKenna's paper was world-opening for me when I first read the early version as a young(er) fellow. Back then, I was primarily struck by their insights on market failures. In IP law, we tend to talk about only one market failure: the fear of copying and the incentive problems generated by the fact that ideas resemble so-called "public goods," which can be easily copied and widely deployed by many users without depleting their value.  Frischmann and McKenna shook my understanding of this assumption. As they write, "[t]hat analysis might be useful, as far as it goes, but it would ignore other market failures, such as the demand-side failure that leads to under-provisioning of drugs to smaller or nonexistent markets." (314).

So, for instance, sure, one reason society may not see enough private investment in developing life-saving drugs is the fact that drug companies could not recoup the cost of their research and testing if they could not exclude generics from copying their novel findings and undercutting their prices. If this is the market failure, then we'd then zoom in on IP rights as the solution to the problem, to prevent that copying and competition. But we could just as well focus on other market failures, such as the fact that some diseases are so rare or so selective in the population they target that there wouldn't be sufficient demand for the life-saving drugs in the first place. All the IP rights in the world would not do anything to stimulate investment if not enough people want or are willing or able to pay for what you're selling. (I also distinctly recall how Yochai Bekler put it in his talk: "it's market failure all the way down.")

But I found new insights in Frischmann and McKenna's article even since the last time I read it. The reason is the extensive work they've done on the final, beautifully crafted version of the article, but also the reading and research I've done in the intervening years. To give just one example of a new insight: I am not even sure I will use the word "innovation" ever again in my work, or at least not without a lot of background or a long footnote defining what the word means.

To give context, many of us might confidently say, to a judge, or a student, or a congresswoman, that the goal of IP, or of any other incentive that seeks to replicate IP (e.g. research grants), is to promote “innovation.” But the reality is that, despite frequent use of the phrase "innovation" in all quarters, from courtrooms to policymaking, the term doesn't have a widely accepted meaning, even within the same field.

Frischmann and McKenna observe that it is tempting for IP lawyers to assume that Article I Section 8 Clause 8 of the Constitution's, the Intellectual Property Clause's, reference to "Progress" refers to one agreed upon objective, or to "brush the issue under the rug" by making the seemingly incontrovertible claim that "IP should promote 'innovation'—as if 'innovation' were one thing." (322).

But it's not. "[A]ny of the following understandings of 'Progress,'" they write, "are perfectly reasonable from an interpretive, historical, and normative standpoint:

A. advancement of the relevant knowledge frontiers—
scientific, technological, aesthetic, cultural, etc. 

B. advancement in the distribution of existing knowledge—
making more of what is known by some, known to all—
framed in terms of education, human capital, or
otherwise 

C. greater quantity of outputs—works and inventions (of
some types) 

D. qualitatively better outputs—works and inventions,
subject to ambiguity regarding the criteria for judging
some outputs “better” 

E. broader participation in creative and inventive
activities—possibly framed in terms of education, human
capital, and/or access to the means of production 

F. increased social welfare, subject to ambiguity about the
meaning of “welfare” 

G. economic growth 

H. sustainable development

(321)  Notably, they put "economic growth" low down on that list (G), and they not unreasonably locate it below "social welfare" (F), even though the dominant theory of why we grant IP rights—the one in the leading case book that I love and use—conceptualizes IP as a fueler of economic growth.

I saw this ambiguity in action recently, in an article I wrote from 2018 to 2020, where I did a review of U.S. state programs in the last few years that used the word "innovation" (or sometimes "technology" or "research") to describe state funding programs. For instance, a program might be called the "Innovation Fund" or the "Innovation Challenge." A person familiar with intellectual property law or federal funding for research might assume such a program would direct financing towards projects relating to bullet point A, advancement of knowledge, particularly scientific or technological; or perhaps towards bullet point F, social welfare, or bullet point G, economic growth.

In fact, these programs had another pretty specific goal in mind: job creation. They directed public money toward projects that promised to bring more jobs to the region than others. That may be a form of "innovation," to some policymakers or even to some voters. It may fall into a subset of bullet point F ("increased social welfare...") or it might even belong as a new bullet point on the list of "Progress" altogether.  But I know for sure that "job creation" is not the same conception of "innovation" or of "Progress" that we use in patent law or IP law. And it's not necessarily the one that the layperson on the street would suspect.

What's more, these disparate ideas of what counts as "innovation" may result in serious conflicts that can undermine a program's efficacy. "Innovation" that promotes advancement of knowledge that also happens to involve automation, for example, would contradict "innovation" that has job creation for people in the region as its ultimate objective. Likewise, "innovation" that promotes advancement of knowledge that happens to involve environmental disruption or the spread of disease, would contradict "innovation" that has sustainable development, social welfare, or longterm economic growth as objectives. Countless other examples of such conflicts could be derived from the list.

In any case, I highly recommend this timely article, even for those who saw an earlier draft.