Why the FDA should legalize Pfizer's COVID pill immediately
The FDA's refusal to legalize Paxlovid and Molnupiravir is killing thousands
The number of people dying from COVID-19 in the United States has been steady at 1,300 per day for the past 30 days. That is equivalent to thirteen 9/11’s each month. Moreover, the number of cases has been starting to slowly tick upwards the past week and there is an expectation this will continue as we go into the winter months and holiday season. On top of this a new strain that many experts believe is worse than delta is starting to spread. The pandemic is not over.
There are two drugs which both have stellar Phase III trial results indicating it can reduce the COVID-19 death rate dramatically - one from Merck (Molnupiravir) and the other Pfizer (Paxlovid). We have new weapons to slay the dragon, but we aren’t using them yet!
We’re in a bad spot right now. Independent Data and Safety Monitoring Boards (DSMBs), operating in consultation with the FDA, have halted enrollment into both trials because the interim results were so good they concluded it would be unethical to continue giving people a placebo. Yet, the FDA has decided to keep both drugs banned. This isn’t because there’s a lot of uncertainty about the safety or efficacy of the drugs, as I’ll touch on below. Nor it is because FDA official’s hands are tied by procedural rules. Both the HHS secretary and the FDA commissioner have the ability to Emergency Use Authorize both drugs right now at the stroke of a pen, but they refuse to act. Blood is on their hands.
The mainstream media isn’t covering this at all. So I was really happy when my tweet on this went viral:
Basic science and data on Paxlovid
Paxlovid is an antiviral drug that inhibits the 3CL protein, a protease that is essential to SARS-CoV-2 replication. It is given in conjunction with Ritonavir, another anti-viral. Ritonavir is not used for its anti-viral activity however, but rather to inhibit an enzyme in the liver that would rapidly break down Paxlovid otherwise.
Pfizer is conducting several Phase 2/3 trials - one is for COVID positive patients at high risk while another is for people who have standard risk (this group includes vaccinated people with breakthrough infections). The third trial looks at the how well Paxlovid can work as a prophylaxis by giving the drug to adults who are at risk of having been exposed to COVID-19 from a family member. Here I’m only going to discuss the results from the high risk group trial.
At the time of the interim analysis, Pfizer’s Phase III trial had 1,219 people enrolled in the trial. The analysis showed an 89% reduction in risk of hospitalization or death for patients treated within three days of symptom onset. For patients treated within five days of symptom onset, the reduction was 85%. No people who received the drug died after 28 days compared to 10 people in the control group.
Bear in mind these results are all from Pfizer’s press release and haven’t been peer-reviewed, so the corresponding caveats apply. It’s worth noting that big companies like Pfizer have a good track record at conducting studies and they have enormous resources to conduct large careful studies that many academics don’t have (Pfizer has invested over 1 billion dollars at-risk to develop and test Paxlovid).
Basic science and data on Molnupiravir
Molnupiravir is anti-viral molecule invented at Emory University around 2018 and developed by Merck in collaboration with Ridgeback therapeutics. It works by introducing copying errors in viral mRNA when the virus is trying to replicate.
The first Phase III trial results on Molnupiravir came from Hetero, an Indian company that has a license agreement with Merck to produce the drug for India and other low-income countries. Hetero released interim results from an N=741 trial way back on 9 July this year. They reported that after 14 days Molnupiravir reduced hospital admissions by 70% (7 hospitalized who got the drug vs 23 in the control group, p < 0.0001).
On October 1st, Merck announced interim results (N=775) from their own study which was performed from more than 170 sites in 23 countries. After 30 days they found that the drug cut the risk of hospitalization by 48% (28 (7%) hospitalized who got the drug vs 385 (14.1%) in the control group). A later follow up with 1,400 patients showed a lower reduction of 30%.
By performing sequencing on 40 of the study participants Merck found that the works equally well against different variants including the delta, gamma, and mu variants. That result confirms what was found earlier in in vitro studies with variants.
The drug’s name is said to come from the name of Thor’s hammer, Mjölnir, which is incredibly bad-ass.
Two additional small Phase III trials on Molnupiravir going on India were ended early due to lack of positive results. Those trials were on patients with
more mild worse COVID-19 (thank you to Josh in the comments for correcting me on this! Based on the data so far it seems Molnupiravir works a lot better in the mild cases rather than moderate ones). There is a concern that at Molnupiravir, because it introduces copying errors in to RNA replication, may lead to the creation of dangerous variants, especially if accidentally given at lower doses. Looking online, there seem to be more experts arguing the risk is very low rather than something we should be worried about. There’s also a theoretical concern that Molnupiravir could introduce DNA mutations and increase cancer risk. Merck did animal studies with large doses looking for this possibility and didn’t see anything, but it remains a tiny risk. Both risks are very small, and the UK Medicines Agency has already authorized Molnupiravir for emergency use on November 4th.
The two pills are complimentary in some ways. The two drugs work via completely different mechanisms, so if given together they can pack a one-two punch (although STAT News suggests that probably neither the companies or FDA will allow that). Also, Paxlovid must be taken with Ritonavir, which messes with liver enzymes and thus has a large number of interactions with other drugs, so it may not be a good option for many people (for instance cancer patients probably won’t be able to take it).
Of the two drugs, Paxlovid is a much clearer-cut case, which is why I will focus on it here. However most of the arguments that will be presented apply well to Molnupiravir too.
The Paxlovid Paradox
Both trials were stopped by a DSMB in consultation with the FDA. Zvi Moshowitz has a great way of distilling the absurdity of the current situation, which economist Alex Tabarrok calls “the Paxlovid Paradox”. It’s worth quoting:
The trial was stopped due to ‘ethical considerations’ for being too effective. You see, we live in a world in which:
It is illegal to give this drug to any patients, because it hasn’t been proven safe and effective.
It is illegal to continue a trial to study the drug, because it has been proven so safe and effective that it isn’t ethical to not give the drug to half the patients.
Who, if they weren’t in the study, couldn’t get the drug at all, because it is illegal due to not being proven safe and effective yet.
So now no one gets added to the trial so those who would have been definitely don’t get Paxlovid, and are several times more likely to die.
But our treatment of them is now ‘ethical.’
For the rest of time we will now hear about how it was only seven deaths and we can’t be sure Paxlovid works or how well it works, and I expect to spend hours arguing over exactly how much it works.
For the rest of time people will argue the study wasn’t big enough so we don’t know the Paxlovid is safe.
Those arguments will then be used both by people arguing to not take Paxlovid, and people who want to require other interventions because of these concerns.
FDA Delenda Est.
Scott Alexander at Astral Codex Ten also wrote about this situation, and in the comments section and on Twitter a number of people raised objections to the FDA EUAing Paxlovid right now. I thought it would be useful to go through them one by one:
Responses to common objections
Chesterton’s fence / setting a bad precedent
Many people on Scott Alexander’s post and on Twitter argue that lowering standards here will set a dangerous precedent long term, and that justifies letting people die here in the name of the greater good. A related argument is that we shouldn’t advocate for removing rules until we fully understand why they are in place (Chesterton’s fence). I keep seeing these objections and they are quite easy to dispose of, so that’s why I’m addressing them first.
The fact is no standards need to be lowered here. None of the people advocating for these two drugs to be un-banned are asking for a full approval from the FDA right now. All we need is for the FDA to use mechanisms that are already in place. The EUA legislation intentionally sets a very low bar for the FDA to act. That’s good because emergencies call for higher risk taking — letting the status quo continue without trying to help is a major risk itself. The more pressing the emergency the more risk should be tolerated. So trying to predict in advance what the right bar should be in the EUA legislation would be dangerous. More generally, bars for approval should always be dynamic, allowing the rational level of risk-taking that the situation requires.
To quote the EUA legislation itself, once a national emergency has been declared the only requirements are:
“.. based on the totality of scientific evidence available to the Secretary, including data from adequate and well-controlled clinical trials, if available, it is reasonable to believe that—
(A) the product may be effective in diagnosing, treating, or preventing—
(i) such disease or condition; or
(ii) a serious or life-threatening disease or condition caused by a product authorized under this section…
(B) the known and potential benefits of the product, when used to diagnose, prevent, or treat such disease or condition, outweigh the known and potential risks of the product..”
(Bolding is mine). The legislation only requires a “reasonable belief” that the benefits outweigh the risk! The way I interpret this is that the legislation is asking that the FDA to do expected value calculation of the benefits vs risk of acting vs not-acting, which is precisely the thing the FDA has shown itself to be horrible at during the entire pandemic. [Note: technically the EUA authority is granted to the HHS secretary, but it has been formally delegated to the FDA.]
All of the specific requirements the FDA have announced for Emergency Use Authorization are things they invented ad-hoc during the pandemic. A lot of people don’t seem to understand this point! The FDA commissioner could rescind those requirements right now and give an EUA at the stroke of a pen.
In addition to EUA there is a second mechanism already in place by which the FDA could make either drug available to at-risk patients right now, Expanded Access. Expanded Access is a special mechanism that allows patients who are at risk of dying to take an investigational drug when there is no comparable alternative. In an op-ed from 7 October clinical trials expert Dr. Marty Makary advocated that the FDA make Molnupiravir available via Expanded Access (he also went on TV on 10 October making the same point). He points out there is precedent for this - the FDA allowed doctors to start using convalescent plasma in certain cases prior to EUA by issuing an expanded access protocol. He also points out that the FDA allowed Trump and other officials to access monoclonal antibodies prior to EUA via this route.
This is the most common objection, and frankly a bit hard to counter since we have no direct data on how many pills have been made and can only make indirect inferences. Note that the FDA undoubtably knows how many are being produced.
Merck says they have been producing Molnupiravir at-risk for months and they are on track to produce 10 million treatment courses of Molnupiravir by the end of 2021 and “at least 20 million in 2022”. It’s not a stretch to infer from these numbers that they already have millions of treatment courses made. Furthermore, since April 27th five Indian manufacturers have been licensed to manufacture Molnupiravir.
I am almost certain that at least tens of thousands of courses of Paxlovid have been manufactured by triangulating from the following facts:
Pfizer has invested over 1 billion at-risk in Paxlovid.
Pfizer says they are on track to produce 180,000 course of treatment by the end of the year (each course consists of 10 pills, given twice daily for five days).
Pfizer already made tens of thousands of doses for their clinical trial.
When the Pfizer vaccine was EUA’d last year, they had millions of doses made.
Derek Lowe says the synthesis of Paxlovid is “much more straightforward” than what is needed for mRNA vaccines and “there should be no reason that the process can’t be run by any country with competent generic drug manufacturers”.
There’s also an important meta-level point which Eliezer Yudkowsky made recently and which economist Tyler Cowen made previously when discussing vaccine supplies (I highly recommend everyone read both). That point is that companies’ decisions about how and when to allocate capital to manufacture and the FDA’s authorization process are not independent. Drug companies and the FDA work in tandem:
If the EUA process were faster and/or a lower bar, in that counterfactual world Pfizer would have been incentivized to ramp up production sooner.
Another issue though is whether the factories will be inspected in time. Hopefully the FDA has already been inspecting Pfizer’s factories and if not will make it their #1 top priority, but sadly I’m not that optimistic. Remember the J&J vaccine doses Biden said the US would donate? Those doses which are desperately needed all across the globe to save lives and stop the creation and spread of new variants? Well, 30-50 million of those doses have been sitting idle in a factory in Baltimore for months, awaiting FDA inspectors. A recent FDA document says that 52 drug applications are currently held up solely because they are waiting for inspections.
The massive inspection backlogs at the FDA have been caused by the FDA shooting itself in the foot by baring their inspectors from flying during the pandemic, delaying 15,000 inspections. It’s a grim situation.
The drugs being made now can save an equal number of lives later
This objection goes along with the manufacturing one. The idea is that supply is going to be very limited and demand will outstrip supply for a while. So, the doses that are being stockpiled now can still be used to save an equal number of lives later after the FDA authorizes them. This objection does carry some weight, and is why I think the net number of lives lost by a Paxlovid delay until December 31st is probably closer to 10,000 rather than 50,000. Analyzing this situation is complicated but consider the following relevant facts:
Pfizer has requested that the first EUA be for the most at-risk only (eg older unvaccinated people a few days into symptomatic COVID).
Between Molnupiravir and Paxlovid, there will probably be enough supply to help the most at-risk starting in January, at least in the US, and generic manufactures will help provide them overseas. The bottleneck constraint will be whether we can identify the people who need it (testing!). Physically getting the drugs to hospitals may also be a bottleneck constraint as well for some time (The companies are prohibited from pre-distributing the drug until the EUA from the FDA).
The marginal utility of each additional dose of Molnupiravir/Paxlovid on the market decreases rapidly. So providing doses now provides much greater benefit than providing extra doses later.
People could/should get the vaccine instead
This objection has come up surprisingly often on Twitter. I believe some people with this objection are not coming from a very good place and are just enjoying making fun of vaccine hesitant people. Personally I have a lot of empathy for people who refuse to get vaccinated - I’m convinced a lot of it just stems from fear of needles, and most of rest from being embedded in the wrong filter bubbles online.
I think some people are mis-applying a libertarian framework which says there’s not much of an obligation to help people for problems they themselves got themselves into. A loosely analogous example would be if someone decides to ride a motorcycle without a helmet and then gets seriously injured as a result and need extensive medical care. Libertarians would say that’s their own fault and taxpayers have no obligation to pay for their medical coverage. Regardless of whether you agree with the libertarian’s argument (for the record, I do not), I don’t really see that it applies here. All we are asking for is that doctors be given the option of giving the drug to people who weren’t smart enough to get vaccinated. So really, libertarians should be in full support of this!
The other thought I think is coming up is people are worried that approving the pill will discourage people from getting vaccinated. There is probably some truth to this second-order effect, but it doesn’t come anywhere close to out-weighing the first order effect of saving lives.
Don’t we already have monoclonal antibodies?
It’s true that the FDA has already EUA’d four different monoclonal antibody treatments and that such treatments are quite effective. However, there has a been a non-stop shortage of “monoclonals” throughout the pandemic because they are difficult to manufacture, because it requires tightly-controlled large scale cell culture which not many labs are setup for. Likewise, we also have the anti-viral drug Remdesivir, but it is not nearly as effective as Molnupiravir or Paxlovid and is also in short supply I am currently visiting my parents in upstate New York, and at the local hospital the wait list for Remdesivir is said to be 30 patients long. There’s also the point that Remdesivir and monoclonals require IV injection in a hospital rather than just being pills you can take at home.
How many people is the FDA knowingly killing?
The fact is, we don’t really know. That’s the thing about lives lost due to FDA delay - we generally don’t know who they are and we often only have rough statistical estimates of how many bodies are in the graveyard (hence the term “invisible graveyard”). Statistical lives may not have a voice, but statistical lives are real lives!
Assuming that the FDA will delay their Paxlovid approval to January 1st, Eliezer Yudkowsky guesses 50,000 will die (Yudkowsky also factors in that the production ramp-up would have started sooner in the counterfactual world where the FDA has an faster/easier EUA process). Zvi estimates 5,000 - 30,000, depending one’s background assumptions.
Here’s an example conservative estimate of the number of deaths from delay on Paxlovid, assuming the delay goes from three days after Phase III study results were released (Nov 8th) until Dec 30 (the Metaculus forecast):
Total time of delay = 53 days
Death rate in US ~= 1,400 per day
Deaths 54*1400 = 74,200
% of patients who get tested and get drug within necessary early window = 25%
% of those patients saved by drug = 89%
74,200 * 0.25 * 0.89 = 16,510
Marginal utility of additional supply at later date (ie Dec 31s- ) rather than earlier - ~50%
lives lost = 8,255
(Note the death rate may drop because Molnupiravir will likely be EUA’d around December 1st, but on the other hand I expect case rates to go up due to the colder weather, dangerous new variant, and holiday visiting, so I’m just assuming the two things cancel out). This estimate also ignores lives that will be lost overseas where generic manufacturers are set to be making and selling the drug.
Is calling it “murder” too extreme?
A number of people (Eliezer Yudkowsky, Zvi Moshowitz, Timothy Carney) have started using much more forceful language to describe what the FDA is doing here, which I applaud. But is the term “murder” justifiable or going too far?
As a starting point, I don't think a person does something that should be called “murder” if they spend money on a vacation rather than saving lives by donating to GiveWell's top charities, even if they are aware that their money would save lives if donated. So I recognize the act-omission distinction has some relevance when judging and talking about behavior.
However, I do think the FDA does do something equivalent to murder when people die because they delay approval of a drug, when the science is clear and when they are in full knowledge that people will almost certainly die as a result of their delay. We need to consider that is in the FDA’s mandate is to protect the public health and when there's a nasty virus going around that means approving life-saving treatments in a timely way. What the FDA is doing right now by delaying Paxlovid is equivalent to a soldier deciding to just sit around and let the enemy attack and kill innocent victims, or a doctor deciding to let a cancer patient die so they can go to the bar and have a drink with friends. People at the FDA have taken on important responsibilities and should be judged accordingly.
So I think the term murder is apt here [legally speaking “negligent manslaughter” is more fitting, but it doesn’t have same rhetorical punch]. Of course, it’s true that FDA officials are sure to have some excuses for their murders, most likely along the lines of "we want to maintain our process / rigor / reputation / prestige". As a proponent of utilitarianism I’m in principle OK with letting people die for the greater good, but only in very exceptional cases!
It’s also worth acknowledging that the FDA does risk their reputation slightly by issuing a quick EUA — the decision may be perceived as them kowtowing to the drug companies or as an erosion or rigor and standards at the agency. But to me, this is like a rich person who refuses to help a drowning child because they are worried about their suit getting wet. The FDA already has an insane amount of rigor and prestige, and an EUA today rather than weeks from now is not going to damage it much. And remember, we are only asking for either Expanded Access or Emergency Use Authorization, not full approval.
The fact remains, FDA officials could sign a document today releasing either drug to at-risk patients and start saving lives tomorrow, but they are choosing not to.
I disagree with the claim that "What the FDA is doing right now by delaying Paxlovid is equivalent to a soldier deciding to just sit around and let the enemy attack and kill innocent victims, or a doctor deciding to let a cancer patient die so they can go to the bar and have a drink with friends."
That is far too generous to the FDA. What they're doing is more like a soldier *obstructing the only exit* so that innocent victims cannot flee as the enemy approaches. Or a doctor deciding to *burn down the pharmacy* so that their patients cannot access the medicine they need to save their lives. Obstruction of life-saving medicine is harmful interference, not a mere failure to benefit.
Don't protease inhibitors delay wound healing and screw up your kidneys?