Correction: Don't Throw Away the Nicotine with the Ashtrays
Evidence on Covid and nicotine leaves open empirical questions about possible benefits
My Simpson’s paradox post has a mistake. Don’t panic: It’s still not Simpson’s. And there’s still no paradox. Keep calm and DAG on.
The mistake was assuming that findings of possible protective smoking-Covid effects reflect only collider bias. Such findings may reflect both collider bias, and a causal effect. The former doesn’t exclude the latter. There’s still a pandemic, Covid costs lives and harms well-being worldwide, and it would be important to know if a widely available drug like nicotine could help protect people from infection or complications, or heal long-term harm. But we don’t know. Science is a social entity made by human beings who make mistakes as a result of common cognitive biases (aka being human). And here it seems likely that confirmation bias (coding evidence as confirming your preconceptions) kept us from running well-designed studies on nicotine and Covid infection including transmission, mortality, and common post-infection problems like fatigue, brain fog, loss of smell, and depression (all of which could be neurological).
This post takes a brief look at the possibility that nicotine may have benefits when it comes to Covid, and examines what it says about science as a social entity that these possible benefits have been widely dismissed (including by me previously) — when the reality is one of uncertainty. With the disclaimer that I’m not a doctor and don’t give medical advice, I do help translate science when people ask me for help, like friends and family often do. So my requested take here is that, if you think you may have lingering symptoms from Covid, you could try nicotine patches for a week and see if they help. We should have better science on this than we do; but you have to see what you can tolerate and what works for you, anyway. Sorry. The world is a mess, and science is part of it.
Review: Smoking Effects As Collider Bias
In winding my way through a few selected highlights of the causal revolution that is slowly replacing the old way of doing science — the old way that talk of Simpson’s paradox and omitted variable bias reflects — I cited Sir David Spiegelhalter and Anthony Masters’s collider bias example from their 2021 book Covid by Numbers: Making Sense of the Pandemic with Data. They say:
the apparent protective effect of being a current smoker… has been a consistent observation in multiple studies and provoked considerable controversy. One possibility is that by controlling for factors that are influenced by smoking, we may be distorting any causal relationship between smoking and Covid-19 risk; in statistical terms, this is known as collider bias (footnote, p. 130).
If you know epidemiology, this rings some bells. For example, it parallels the case of alcohol and cardiovascular disease. Studies long found protective effects for alcohol in moderation. But it turns out that, probably, alcohol is just so poisonous that sicker people disproportionately self-select to not drink at all — thus distorting health correlates through selection bias. Health predicts health. Illness predicts illness. Selection bias sends a ripple through these endogenous selection effects when illness leads to less unhealthy behavior among the unhealthy.
So I didn’t dig into the idea that, if smoking looked protective, it was nothing more than collider bias, a disturbance in the statistical force. Smoking and drinking are bad for you. Very bad. Sicker people therefore do them less, to avoid the damage they can’t afford. And so, in observational data, these behaviors may look beneficial. But that’s silly and we know better.
Going down a layer into methods, Spiegelhalter and Masters continue “A model that only adjusts for demographic factors… showed a positive link between smoking and death from Covid-19.” To be fair, this was just a quick explanation in a footnote. The authors don’t say that there is no possible benefit of nicotine in the context of Covid, or that we should as a rule resort to simple, demographic-focused modeling to prevent collider bias.
We shouldn’t. Stripping models down to demographic covariates as a first-line defense against the possibility of collider bias would be dangerous. Demographics are among the most political variables imaginable. Race, gender, age — you name it, it’s fraught with power. Power factors into a lot of classic selection bias problems (formerly known as Simpson’s paradox examples), like racial inequality in criminal sentencing and gender inequality in admissions and wages. Whether models are simple or complex, we want to draw causal diagrams before running analyses to check for colliders, lest conditioning on a collider accidentally introduce more bias than it corrects for.
Again, methodologically, the point is not that an apparent protective effect of smoking on some Covid outcomes is not a good example of possible collider bias. It is. The drinking analogy holds. If you ever hear that something unhealthy looks like it’s maybe a little good for you in observational data, you should think “this could be selection bias, because unhealthy behavior is probably unhealthy.” There is overwhelming evidence that smoking is unhealthy.
But the logical point is that possible smoking-Covid benefits can illustrate collider bias in models that didn’t use causal diagramming (bad science), and nicotine may have Covid-protective benefits. The former does not preclude the latter.
Maybe smoking kills, and smoking with Covid kills harder. But maybe nicotine has benefits against Covid, too. What do the data say?
Daily Non-Neutrality Reminder
This is your daily reminder that neutrality is a myth. Data don’t collect, analyze, and interpret themselves. If we programmed an AI to do that for us, it would still be an artefact of its creators; in Evans et al’s parlance, we don’t collaborate with what we design.
In scientific discourse including risk communication, this means, as Greenland quips:
“DATA SAY NOTHING AT ALL! Data are markings on paper or bits in computer media that just sit there… If you hear the data speaking, seek psychiatric care immediately!”
This critique of the left-liberal “follow the science” mantra carries from the upper echelons of methodology (Greenland) to popular science Youtube (Hossenfelder) to intersections of critical science and popular science (Prasad). But while this countermelody is well-known, the mantra itself is still a popular dogma. And it can be hard to break through to adherents in a single conversation with the message that this reflects a misunderstanding of science as secular religion, it’s not, and we don’t know most of what we’d like to know, including what it is that we don’t know.
So non-neutrality (bias) keys into overconfidence (hubris). The solution must be twofold: We can have better (more neutral, accurate) scientific information. But we also need more humility, since neutrality is impossible. People tend to dislike this, because uncertainty is hard. No one likes to be criticized, to admit they were wrong, or to preemptively warn people from a position of power that they could be mistaken. Eating humble pie and performing expertise don’t mix, at least not in most current social and professional norms.
Covid and Nicotine
There’s a lot of Covid science out there, and much of it got there fast. Some of it is good science. But the Covid publishing bonanza came at a cost to openness, diversity, and quality in science. Misinformation is rampant all-around, including in consensus Covid science communication. In the nicotine sub-corridor, we wound up with a bunch of inconclusive evidence linking nicotine with possible Covid-related benefits. (Much of this section repeats things I’ve already Tweeted trying to get someone else to give me the answers so I wouldn’t have to write this post, which mostly says I couldn’t find and don’t think we have the answers.)
Don’t go straight for the Nicorette. Nicotine is possibly not good for you all by itself as a habit, and degrading your baseline health is a bad idea during a pandemic and generally. For instance, nicotine may cause erectile dysfunction; and things that are bad for your dick are usually bad for your heart (e.g., sedentary lifestyle, stress, alcohol). Thus it’s unsurprising that “Nicotine may contribute to cardiovascular disease, presumably by hemodynamic consequences of sympathetic neural stimulation and systemic catecholamine release.” There are probably better options for whatever you would do nicotine for as a psychogenic substance, anyway (big topic).
In terms of Covid benefits, apparently, the theory is that nicotine displaces Covid on receptors. So chronic use isn’t necessary. This theory has face plausibility if you know anything about nicotinic acetylcholine receptors, which featured prominently in Jeremy Narby’s ayahuasca book, The Cosmic Serpent: DNA and the Origins of Knowledge. There, these receptors are presented as a kind of cosmic radio transmission receiver in the context of psychedelics and shamanic knowledge. Here, we might care about this because it makes them seem both central to consciousness and manipulable. “The more you give nicotine to your neurons, the more the DNA they contain activates the construction of nicotinic receptors, within certain limits” (p. 119). In this story, nicotine primes the nicotinic receptor pump.
So one might envision three sorts of Covid-nicotine effects in a purely mechanical sense: Nicotine could hog these receptors, protecting them from Covid infection. It could kick Covid off the receptors. And it could create new receptors, even if it couldn’t kick Covid off. So that’s a theoretical basis for possible Covid-nicotine benefits, but what does the evidence seem to suggest?
Maybe long Covid is “a severe impairment of acetylcholine-orchestrated neuromodulation that responds to nicotine administration,” according to Marco Leitzke (2023 Bioelectron Med). “Several investigators could demonstrate that the SARS-CoV-2 related spike glycoprotein (SGP) attaches not only to ACE-2 receptors but also shows DNA sections highly affine to nicotinic acetylcholine receptors (nAChRs).” So maybe Covid and nicotine are competing for the same receptors, and that competition could be useful.
Covid-Nicotine Misinformation
We don’t know a lot of what we’d like to know about Covid and nicotine. That didn’t stop a lot of people from panning the possible benefits, though. For example, Jamie Hartmann-Boyce of the University of Oxford Centre for Evidence-Based Medicine called the evidence “weak and contradictory,” concluding:
we are unlikely to know whether nicotine replacement has a role in COVID-19 any time soon.
In the meantime, there is no value in people purchasing nicotine replacement to help protect themselves against COVID-19. Such a move could cause harm by reducing the availability of nicotine replacement therapy for people who wish to quit smoking. For now, nicotine supplies must be preserved for the people who need them.
Notice the leap from not knowing if there’s a benefit, to there being no value in trying to get the possible, because other people might need the supplies. This takes the following steps: We don’t know → It doesn’t matter that there’s a possible benefit, because supplies are limited. → Other people are more deserving of a certain benefit than you, the reader of this science communication, are of the possible benefit.
This is reminiscent of when the U.S. CDC told people to not mask early on in the pandemic, in a probably nobly intentioned lie to keep limited protective gear supplies for healthcare providers. But this is empirically and ethically wrong. Just because there is a “better” (value judgment) use for limited resources, doesn’t mean those resources don’t have potential value in multiple contexts. Science communicators should not be in the business of lying, nobly or otherwise. Nor of making value judgments for the people whose choices they’re supposed to be informing.
The value judgment in both cases here is utilitarian. It assumes that we should all agree — or have agreed — to use available resources for the greatest good for the greatest number. Then it doesn’t do out the calculations to prove that these choices would really have those effects (an open question).
Public health institutions shouldn’t lie to some people about masking during an airborne pandemic to help others. Evidence-based medicine centers shouldn’t lie to some people about possible value in having nicotine on hand in case it helps fight the virus in order to help others, either. People should get to make their own choices based on the best available information in free societies.
Intentionally constraining those choices through lying also probably reflects poor strategic judgment, because it degrades trust when people see what you’re doing. But we’re all swimming in so much misinformation, so much of the time, that maybe it nets out sometimes. Because normal is just so bad.
CEBM put this out back in 2020. The CEBM website, too, still links to this 2020 synthesis of nicotine-Covid evidence. What’s the current state of the evidence?
A Brief Tour of Some Covid-Nicotine Science
My searches haven’t turned up the randomized trial of my dreams. For instance, PubMed returns zero results for “transdermal nicotine long Covid trial.” This may seem surprising, if you expect science to operate rationally (unlike its human actors). After all, we knew early on in the pandemic that there was promise, a lot of people have spent a lot of time and money doing and publishing Covid research, and we have the tools to figure out if this lead goes somewhere. Didn’t we?
Apparently not. This systematic review to design Covid-nicotine trials (Dautzenberg et al, Respir Med Res 2021) might be useful for its specifics: 3.5 mg nicotine/day patches, increasing gradually up to 14-15 mg/day over two weeks, using 24-hour patches, and decreasing gradually over 21 days. It seems odd that this article figured out how to do Covid-nicotine trials in 2021, but has only been cited twice since then — and only one of those citations published results from such a trial.
That review’s trial results citation was Labro et al (Intensive Care Med 2022), reporting results of a trial using nicotine patches on Covid patients on ventilators. The trial wasn’t big enough to say whether or not there was a mortality effect. But that didn’t stop researchers from wrongly claiming “Day-28 mortality did not differ between the two groups (30 [28%] of 106 patients in the nicotine group vs 31 [28%] of 112 patients in the placebo group; odds ratio 1.03 [95% confidence interval, CI 0.57–1.87]; p = 0.46).” According to these data, it’s possible that nicotine benefited survival. The published interpretation saying otherwise is an example of statistical significance testing misuse. It’s hard to do a PubMed search without finding this common mistake. Scientists rose up against it, but journal editors didn’t get the memo, or something.
Anyway, the evidence here remains weak and inconclusive, just like the CEBM website says. It’s just also still possible that there’s a benefit. And if there’s a small life or death benefit, that would be important to know. So we shouldn’t say that we know it doesn’t exist. We don’t know that from this evidence. Uncertainty aversion may be hazardous to your health.
That’s about survival in already-severe cases. What about neuroprotection? According to Letsinger et al, “Nicotine exposure decreases likelihood of SARS-CoV-2 RNA expression and neuropathology in the hACE2 mouse brain but not moribundity” (Scientific Reports, Vol. 13, No. 2042, 2023). Specifically, “pre-exposure to nicotine decreased the likelihood of SARS-CoV-2 RNA expression and pathology in the brain.” In other words, nicotine exposure exerted neuroprotective effects in mice. Why we were studying this in mice in 2023, beats me.
In Praise of Citizen Science
It would have been a no-brainer in 2020 to design a simple 2x2 experiment people could volunteer for using nicotine patches to prevent and treat Covid, and to treat long Covid. This could still be done tomorrow using a simple online platform. Maybe there is more of a role for this type of citizen science in the current psychosocial ecosystem, than infrastructure currently supports. Because science has problems, and society does, too. But people are perfectly capable of opting to roll the dice and experiment.
So why not let them do that within an open science framework? Otherwise they’ll experiment and we won’t get the data. Or not experiment and suffer. Doing more and better science to at least try to help prevent suffering seems like a no-brainer. Not doing it is (on the face of it) a fully solvable infrastructure problem.
I wrote about this idea previously in the context of abortion myths, suggesting that there are a number of simple experiments women could opt to participate in without much centralized administration to test possible treatments for common and potentially serious post-abortion mental health problems (e.g., progesterone, iron, and folate supplementation).
No one is doing them because no one is taking seriously both the evidence on harm (threatening to pro-choice interests) and the possibility that abortion still benefits many women while harming others (threatening to pro-life interests).
And perhaps nobody who’s interested in taking both these things seriously as a matter of science (not to mention feminism [and/or humanism]), wants the headache of upsetting both groups. But it really matters for women’s health that we do. One way to run experiments that are too legally tenuous to run centrally, is to collect data from people doing their own, decentralized Internet self-administration, like James Fadiman and Sophia Korb’s microdosing work. This might be a good method to work around institutional sexism to advance women’s health research. Progesterone is over-the-counter in the U.S., iron and folate are available everywhere, and these experiments could be done.
Fadiman and Korb’s model ports to the nicotine-Covid context, too. The logistical question is whether a general-purpose citizen science platform could enable people to run lots of different surveys and experiments on this model; or whether it takes small teams of dedicated, platformed researchers to do a little good work in each subarea. The former looks like an important possible advance in citizen science, while the latter looks like (maybe) scientists dealing with collapse and corruption in scientific institutions by exiting when they can/must (an escape valve from perverse incentives worsened by administrivia).
Both ring true. So one could envision trying both incarnations of this type of platform, iterating and combining, and see what works.