Smoking and Covid

Paul Kedrosky wrote:

This paper is getting passed around today, with its claim that there not only isn’t a causal relationship between smoking and COVID, but possibly a protective role. This sort of thing drives me crazy about pre-prints. If your data suggests a conclusion that runs counter to decades of prior work with better data — in this case, that there is a strong relationship between smoking and lower respiratory infections — you should always consider the possibility that your new data is crap, not that instead of x → y you have somehow proven y → x.

I replied: Sure, but it’s fine for them to post the research, right? People might be overreacting to it, but the article itself seems clear enough.

Kedrosky responded:

It’s fine to post the research, I guess, but I don’t see what we learn from it, other than that the underlying data is flawed. I would rather see a paper saying, “You want to know bad China’a COVID data is? I can use it to prove statistically that cigarette smoking protects against lower respiratory infections”. That’s a better paper.

That makes sense, but you have to be open to the possibility that smokers are less susceptible to the disease, right?

P.S. I put this under the 6-month delay because, yeah, coronavirus is topical, but this smoking-and-coronavirus thing is not such a big deal. And in 6 months maybe people will want to read some retro coronavirus content. So here it is!

106 thoughts on “Smoking and Covid

  1. In September 2020, this meta-analysis found that smoking is a risk factor for more severe COVID-19 outcomes: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7239135/
    16 of the 19 reviewed studies came from China, which I think casts doubt on this peculiar (to me) line of reasoning from the above post that “if this study using Chinese data found that smoking is protective, then Chinese data is terrible”. Meanwhile the preprint from this post is still unpublished, FWIW.

  2. I didn’t check this paper yet, but the protective role has been known since the first SARS in 2003.

    Missing smokers for two similar viruses 20 years apart but not in between, how do you get that from selection bias? And the effect is so huge and obvious that there must be multiple mechanisms. There are probably adaptations to chronic respiratory tract inflammation and low blood oxygen along with some effect on ACE2.

    You can find dozens of papers before May 2020 on this here:
    https://old.reddit.com/r/COVID19/comments/faluhv/an_exhaustive_lit_search_shows_that_only_585_sars/

    • Anon,
      All ‘smoking’ papers do is notice that hospitalized are usually not smokers, not that smoking prevents hospitalizations. I would ascribe that to statistical noise. You could as well look at people who eat more than three bananas a week and their relationship to hosp.

      Reality is that around 80% of ‘cases’ have been asymptomatic and there are all sorts of people with preconditions among those 80% who did not get hospitalized. Obesity/diabetes may make things worse, but there are many who are obese and such that did not get hospitalized, etc. Same with other conditions.

      Research needs to start with all infected people compared to hospitalized (that would be very difficult as inappropriate tests have been used and we have no clue who and how many have been exposed).
      Instead, studies like the one above start with hospitalized and then reverse engineer things (20/20 hindsight).

        • Isn’t the paper assuming that the rate of infection in smokers is the same as that in non-smokers? And, if smoking was considered a risky behaviour for getting very ill with covid-19 then smokers who act rationally are going to change their behaviour to avoid catching covid-19. It’s like younger people are much more likely to be in hospital then previously because their behaviour is much more risky then older people’s is now.

        • The question remains, why is this only happening for SARS and SARS-2?

          Unless your theory explains why this amazingly consistent phenomenon shows up only for SARS and SARS-2, then you can dismiss it.

        • I don’t understand why coming up with a “theory” is so important to you. There’s numerous possible theories.

          1. Perhaps patients heard news about a dangerous mystery respiratory illness, and thus feel that by suppressing any mention of smoking they would allow their coughing to be treated more seriously than doctors
          2. Perhaps unlike prior illnesses, coughing is a more prominent symptom and thus patients are more likely to cease smoking, and backdate their cessation of smoking. (Note that the French study had 32% be former smokers)
          3. Perhaps there was a recent anti-smoking drive in the areas being sampled, and thus the patients chose to avoid the social stigma of smoking, which they felt might lead to their activities being blamed for their ailment
          4. Perhaps as was in the case of the chinese study, over 90% of patients failed to smoke their smoking status because they were too ill too communicate, which meant that the ones that gave their smoking status were disproportionately the less sick patients.

          You can come up with hypotheses all day long. You still haven’t established that there’s even anything there yet.

    • They failed to do exit polls but I doubt the blinding was very effective anyway due to the side effects.

      But failing to test 477 suspected cases when you only had 170 positives in the trial looks pretty bad. There were also 4x more subjects excluded for “other protocol deviations” in the vaccine group (311 vs 61). Apparently the FDA didn’t see any need for further explanation so we don’t know what that means.

      https://www.fda.gov/media/144416/download

      • Long is right! Maybe this article has concrete discussion of the original factual claims. I don’t know. I gave up after the n-th paragraph of hand-wringing over the BMJ giving ammunition to spooky twitter anti-vaxxers, snippets of other pro-vax twitteratti dunking on the former and of course endless discussion of “framing”.

        > I will also preface the remainder of the post by emphasizing that, if Thacker’s report is accurate, there did appear to be some significant problems with the three sites in Texas managed by Ventavia

        This was at least a helpful warning the subsequent 5000 words would be orthogonal to the actual factual accuracy of the claims.

    • Once again, they need to explain how this collider bias shows up only for SARS and SARS-2 but not the flu, etc.

      It is also quite peculiar that covid shares symptoms with high altitude sickness, which smoking also turns out to help.

      • Anon,

        It could be because flu affect way more people every year (more heterogeneous and representative sample) than SARS and SARS-2, so your ‘effect’ disappears.

        I am not arguing one way or the other. It was known among high performing high altitude athletes that smoking somehow helps in oxygen-deprived environment, etc. However, proper comparisons with infected/not hospitalized needs to be performed, even though it is next to impossible, in order to find a causal link.

        I think it only ‘helps’ to the extent that smokers are used to oxygen-deprived environment, they don’t react as much as others to low O2 saturation, but it doesn’t mean it helps them not to get COVID-19 in the first place

        • It could be because flu affect way more people every year (more heterogeneous and representative sample) than SARS and SARS-2, so your ‘effect’ disappears.

          It wasn’t seen for MERS either. Or heart attacks.

          Just SARS and SARS-2, in every country, seperated by ~20 years.

        • You don’t really need to come up with a specific explanation, there’s no consistency in the frequency of self-reported smoking prevalence amongst flu patients. The best evidence is that hospital self reporting is extremely variable depending on context, location, demographics etc – and thus conclusions based on comparing the number of smokers at a small number of hospitals in one specific city in a specific pandemic context to population level surveys across the entirety of China is extremely perilous.

          This paper is from March 2020 anyway. Someone tried to do a trial using nicotine to treat COVID-19 using this hypothesis. Given the trial apparently completed in april this year and despite lots of initial fanfare, the researchers have gone completely silent, I suspect that unsurprisingly, they didn’t find anything. https://clinicaltrials.gov/ct2/show/NCT04598594

        • As I pointed out the previous time we had this discussion, the important point is not to find a flu paper with low smoking rates, the important thing is that of the flu (or other hospitalization with reported smoking numbers) papers you did find, all of them gave very different numbers for smoking prevalence. The case that absent a causal explanation, hospital self reports should be close to the national survey figures is very poorly made. That’s what you need to prove, not some specific thing about flu.

        • I remember you found a paper that included a bunch of children with the flu then compared their smoking rates to adult smoking rates.

          Is that the paper you are thinking of? Maybe there were more.

        • I don’t remember that or see the relevance. I am sure it was based on actual data though.

          Anyway, you have no sources to back up your claim that smoking rates in flu patients are wildly inconsistent with the smoking prevelance.

        • Yes, one of my randomly searched for flu papers included a study with children in it, but the entire point I was making was that the population choice and context makes a difference to proportion of people reporting smoking. Like, I wasn’t trying to prove somehow that flu also has a correlation with low smoking! Like if you accept the importance of the confounder age on this flu sample, why do you reject adjusting the analysis of SARS smokers for health professional status? Why is it legitimate to adjust *only* when it helps your conclusion?

          What *you* need to show is that there is a reasonable expectation that absent a direct medical causal explanation, hospital self reports in this kind of pandemic context should be accurately related to national level survey data. This alternative where you go nah-nah to any explanation because it subjectively doesn’t make sense to you, and also when any positive explanation of the finding fails replication you switch to another one, is clearly fruitless.

          If I did find another illness with reduced smoking levels, I assume you will just say that this shows that other illness has some hitherto unknown similarity with COVID-19. (Ironically, if your theory is that it’s not ACE receptors but something to do with direct physical action on the lungs, then it’s strange that you *don’t* see the same effect on other respiratory diseases that have very similar symptoms.)

        • Yes, one of my randomly searched for flu papers included a study with children in it, but the entire point I was making was that the population choice and context makes a difference to proportion of people reporting smoking

          I only remember the one flu paper, if there were others they would have been incorporated into the reddit post linked above. So I suspect you are misremembering.

        • > I don’t remember that or see the relevance. I am sure it was based on actual data though.

          The relevance is that medical workers, who contributed the vast majority of SARS patients in that study, *are not allowed to smoke*.

        • The relevance is that medical workers, who contributed the vast majority of SARS patients in that study, *are not allowed to smoke*.

          I’m not certain which study you are referring to here (but I suspect I know). Either way, now there are hundreds of other studies reporting the same so that explanation looks rather silly.

        • Navigator-

          > Instead, studies like the one above start with hospitalized and then reverse engineer things (20/20 hindsight).

          The problem ​(imo) is satisfaction with a jump from correlation to causation without testing that causation in obvious ways (e.g., looking for a dose-effect) or doing a robust investigation of theorized causal mechanisms.

          Not to mention, a near total lack of satisfying basic requurements – like that the sampling is meaningfully representative.

    • Replicating the results is so important, in this case. For those who haven’t handled data that can’t be verified, there are many reasons for this collider bias (is there, for instance, a biochemistry test that the person is an ex-smoker or is it just taken for the patients’ word that they are).

      Normally, this forum is somewhat enlightening. This thread is bizarre. Finding reliable data, verifying it is, and replicating the results is what’s needed. Everything else is semantics, in my opinion.

      • The results have been replicated over and over, and are consistent far beyond any other type of observational healthcare data I’ve ever seen. That isn’t the issue.

        The issue is why this happened ~2003 for SARS, then went away, then returned for SARS-2. That is not how “statistical noise” behaves.

        • What do you think is hyperbolic in that comment?

          This is literally the most consistent effect I have ever seen come out of observational healthcare data. The norm is a huge pile of conflicting studies. And then to clinch the matter, it appears and disappears along with SARS viruses.

    • This is such an evergreen citation.

      It got me thinking, this paper probably won’t get its due appreciation because its main impact will be in preventing bad studies from occurring in the first place, which doesn’t lead to citations. Its only citations will be in papers explaining why bad studies were bad.

  3. > Sure, but it’s fine for them to post the research, right?

    I think there oughta be a law that with this kind of paper, the authors have to address Hill’s Criteria for causation, and in this particular case whether there is a “biological gradient.”

    Seems to me that if you think there’s a casual mechanism whereby smoking reduces the severity of COVID, you’d immediately think it’s important to examine whether there’s a signal in your data of how much people smoke. Wouldn’t it be kind of nuts to think that if there’s an effect, you should treat someone who smokes two packs a day the same was as you treat someone who smokes maybe a pack a week?

    It’s fine to publish an article that says that in a particular sample, “current smoking” wasn’t associated with hospitalization.

    But if you want to say this:

    > This preliminary analysis, assuming that the reported data are accurate, does not support the argument that current smoking is a risk factor for hospitalization for COVID-19, and might even suggest a protective role.

    then, IMO, you need to dig deeper. There evidence isn’t really sufficient, IMO, to be characterized as “not supporting” the argument that current smoking is a risk factor for hospitalization for COVID, let alone as suggesting a protective role.

  4. I don’t know anything about this issue, but I note that even if we accept the claim that “there is a strong relationship between smoking and lower respiratory infections” this doesn’t mean smoking can’t be protective against COVID.

    For one thing, it’s my understanding that COVID often initially infects the nasal passages, so it’s not necessarily a “lower respiratory infection” at all. Second, not all respiratory infections (and infectors) are the same; perhaps COVID is different in some important way. I gather some plausible mechanisms have been suggested.

    There’s plenty of evidence that smoking is terrible for you in a lot of ways, so I think it’s appropriate to be skeptical of claims that there are any health benefits at all…but not so skeptical that such claims should be dismissed out of hand.

    • For sure, but nevertheless the core methodology of comparing specific hospital self-reports to population level survey and assuming causation on that basis is *extremely problematic*. I think more clearly damning is the (AFAICT) failure of nicotine to be effective in double blind trials. (As in, I’m aware of at least one trial on treating COVID with nicotine that ended in April, and the total lack of any reporting on its results suggests strongly to me that they failed to find any positive result.)

    • “but not so skeptical that such claims should be dismissed out of hand…”

      Yep.

      American & British tobacco companies didn’t invent smoking. People of all cultures all over the world smoke, they smoke all kinds of different things and they have been smoking for millennia. It’s hard for me to believe that such a universal behavior has no beneficial effect. I’ve seen claims by anti-smoking advocates that people who quit before age 40 – approximately the expected human life span only a few centuries ago – have virtually no long-term health effects from smoking.

      Radiation isn’t good us but people with cancer are often treated with it. In the same way, while smoking has negative effects on humans, the negative effects on pathogens within humans could be much worse.

      So I think this is interesting. I really want to know: why do people everywhere smoke?

        • Sex has a pleasurable sensation but it serves a function as well.

          Just looking around the internet quickly I don’t find much authoritative information but it appears that many different compounds were used in American indigenous smoking, so probably “smoking” in general is just one way to access active compounds in the plant world, and may have been used for a variety of different compounds and effects, both psychoactive and medicinal.

      • most of the things that are smoked are psychoactive and physically addictive. tobacco in particular, still the most-smoked thing, has mild stimulant effects that are pleasant to many people, and more insidiously creates the false impression among addicts that cigarettes calm them down when in fact what happens is the addiction makes them feel shitty and anxious after too long without a cigarette, and satisfying the craving improves upon that baseline (but likely NOT upon the no-tobacco baseline for most users). it really doesn’t seem more complicated than that.

        the fact that people still believe otherwise after all this time really goes to show how effective industry propaganda was at muddying the waters around the state of the research.

        • “tobacco in particular, still the most-smoked thing, has mild stimulant effects…”

          Yes, this is a good point. Although it’s surprising you say they are “mild”. If you need to stay awake for many hours at a stretch with legal drugs, in my experience tobacco is far more effective than caffeine.

          “creates the false impression among addicts that cigarettes calm them down….”

          As Phil’s quote suggests, this sensation isn’t restricted to addicts. Almost everyone who smokes says that it’s “relaxing” – but I think what’s going on is that they’re tired possibly from being stressed out and the nicotine gives a boost that erases the tiredness.

          None of this means however that tobacco or smoking in general doesn’t some beneficial effect, even if over the long term the overall effect is negative.

        • jim –

          > None of this means however that tobacco or smoking in general doesn’t some beneficial effect,

          I’m not sure what that means. We engage in all kinds of behaviors that have clearly negative impact long term. The measurement of “benefit” seems very complicated to me.

          As for smoking in particular….I smoked for many years. I would never have described it as providing a “benefit.”

          But more to the point, I developed the habit when it was a positively miserable experience. It make me cough. Made me feel sick. I had I force myself to inhale. Nothing pleasant or even remotely “relaxing” about it. But I was about 15 and I thought it made me cool. So the “benefit” was an entirely subjective assessment – and something that could be applied to any of many, many behaviors that would be kind of hard to objectively describe as beneficial.

          Once I had developed an addiction, the “benefit” was more of a rationalization I told myself to justify the habit despite the expense and health risks.

        • Maybe you’re right, humans have been engaging in this harmful activity for tens of thousands of years completely unaware that it’s entirely detrimental.

        • Jim:

          Maybe not completely unaware. It makes you cough, right?

          One thing I learned a few years ago from reading a book about cigarettes and smoking is that modern cigarettes are designed to make it easy and comfortable to smoke. Apparently, raw tobacco can be irritating to the throat, so you can’t easily smoke the equivalent of “2 packs a day” of hand-rolled. The point is that smoking, as practiced in the mid-to-late twentieth century in this country, was much worse for your health than smoking had been in earlier centuries. Put briefly: “cigarettes” are a lot more dangerous than “tobacco.”

        • jim –

          I just think it’s a little complicated.

          Partly because the interaction between “pleasureable” – and in particular pleasurable in the short term – and “beneficial.”

          And then there’s the (somewhat tangential?) connection to the complicated relationship between pleasure and pain – e.g., BDSM.

          Since you like Wright, here’s a recent podcast where he and Paul Bloom discuss some of that (again, they’re both generally too facile with the whole ev psych aspect for my tastes, but interesting nonetheless).

          https://bloggingheads.tv/videos/62998

        • Andrew said:

          “The point is that smoking, as practiced in the mid-to-late twentieth century in this country, was much worse for your health than smoking had been in earlier centuries”

          That’s a reasonable speculation. Expanding on that: readily available cheap manufactured cigs in 20thC drove up the frequency of smoking dramatically. Say you decided to go to CA for the gold rush in 1850. Just about any way you went it took months and there was no corner store to buy cigs on the way. Whatever the negative or positive effects, you’d have to concede your habit or carry a half ton of tobacco with.

          “It makes you cough, right?”

          Some people cough frequently when they smoke but among all smokers in my experience coughing is an infrequent event. From what I’ve seen probably the most universal readily visible negative health response to smoking is the phlegm it causes – smokers are always horking. :( Gross! That’s the thing I hated most about it.

        • Joshua said: “I just think it’s a little complicated.”

          Me too! Humans are complex. But humans are also animals subject to selection and severely deleterious behaviors are strongly selected against. I think it would be very unlikely that pre20C smoking is survival negative. However in 20thC surely people smoked dramatically more than in the past.

          I think the psychological effects of smoking have been dramatically under-studied. I suspect that, like caffeine, the stimulant aspect of nicotine is beneficial for brain function. Im not quite sure where to put your thoughts on BDSM, but should we call it a universal human behavior? I’m hesitant to do that.

    • I agree and don’t find this particularly surprising, in the same sense that amputating your leg “protects” against MCL injuries.

      There is also a negative association between smoking and weight which could confound things.

  5. How difficult can it be to replicate the results of this study? After skimming the paper, it seems that all that’s required is a) reliable data (not from China, say) b) use a sampling algorithm such as MCMC. Then, because the p-value is so low, the sampling distribution should easily converge to a stationary distribution.

    Right? Am I missing something? If this is true, than there should be very few smokers who are hospitalized with Covid-19 — everywhere in the world.

    PS Really, Andrew, you need to fix the design of the site. It’s terrible.

    • Once again, the explanations offered here do not explain what we see. Eg, they totally ignore that the same was seen fot SARS. They also conflate p(severe covid | smoking, covid+) with p(severe covid | smoking) and p(covid+ |smoking).

      These are really basic errors being made over and over. It is insane to get health advice from such people.

      • Smoking is known to increase the risk of infection of both bacterial and viral diseases, such as the common cold, influenza and tuberculosis1, and smoking is a putative risk factor for Middle East respiratory syndrome coronavirus infection2.

        Looks like they did a literature search and totally missed SARS and smoking, which you would think is the first thing to check. Or they didn’t like what they found so totally ignored it and went on to produce a dishonest document.

        Incompetence or fraud, the end result is the same.

        • Anoneuoid –

          > Incompetence or fraud, the end result is the same.

          Was it incompetence or fraud that has led you to, repeatedly, make simple and obvious mathematical errors (so simple they were obvious to me) in ways that minimize the magnitude of your theoretical errors?

          Or I guess I should ask whether it was incompetence of fraud that caused you to “ignore” the data that made your obvious errors look worse?

          Anoneuoid
          on October 24, 2021 at 9:27 pm said:

          […]

          Weekly cases are at ~500k right now, so I was off by 2.5 weeks. They should start rising soon.

          Actually, cases for that week were over 540k. We didn’t reach ~500k until some 5 weeks later. Cases still haven’t started rising yet.

          You made that statement after having previously made incorrect errors, because of your failure to account for the lag in reporting. And I informed you of that error you were making.

          On September 24th, you underestimated the per day deaths in Florida by somewhere in the range of @50%-100%.

          https://statmodeling.stat.columbia.edu/2021/10/22/how-did-the-international-public-health-establishment-fail-us-on-covid-by-explicitly-privileging-the-bricks-of-rct-evidence-over-the-odd-shaped-dry-stones-of-mechanistic-evidence/#comment-2026919

          And that was after your error of not accounting for the lag had been explained to you.

        • I thought I responded to this in an earlier thread, but you keep posting these silly off-topic “gotchas” so will do it again:

          Actually, cases for that week were over 540k. We didn’t reach ~500k until some 5 weeks later. Cases still haven’t started rising yet.

          First, 540k ~ 500k here. While we needed a precise number and date for the bet, this type of data doesn’t really allow that kind of accuracy.

          Also, I see the week ending 10/23/2021 is currently 505,350 cases according to JHU (covidData R package) and 474,686 cases according to the CDC data.

          So I have no idea where you got your 540k number but it is off by 10-20% from the others, probably due to different ways of handling reporting lags, etc.

          On September 24th, you underestimated the per day deaths in Florida by somewhere in the range of @50%-100%.

          Here is the original post:

          Apparently Florida schools opened on Aug 10th when there was about 20k cases per day, after 1.5 months of no mask mandates there are now under 10k cases per day. At the same time deaths/day have dropped from about 200 to 50:
          https://www.worldometers.info/coronavirus/usa/florida/

          Meanwhile NY has a statewide mask mandate and their covid cases have stated flat at ~5k per day while deaths have tripled from ~13 to 40.

          I see the updated count is currently 192 for Sept 24th, down from 278 on Aug 10th. The case count has not changed much, and NY data is still about the same too:
          https://www.worldometers.info/coronavirus/usa/florida/

          So we see that as of Sept 24th, the state with mask mandates has pretty much flat (actually slightly increased) cases and increased deaths, while the one without has decreasing cases and deaths. This pattern has continued to today.

          Finally, nothing about the above indicates any type of fraud or incompetence. Making wrong predictions is not a bad thing in science, repeatedly failing to include SARS-smoking data in the literature reviews on SARS2-smoking does indicate something is wrong.

  6. David Simons, Lion Shahab, Jamie Brown, Olga Perski. (2021). The association of smoking status with SARS-CoV-2 infection, hospitalisation and mortality from COVID-19: A living rapid evidence review with Bayesian meta-analyses (version 12). Qeios. doi:10.32388/UJR2AW.15.

    Most comprehensive review available so far.

    • +1
      The comment section (or maybe just Anoneuoid, who comments a *lot* on this blog) seems to immediately forget about single studies being contingent values of random variables, at best

        • I haven’t seen where you “cited” dozens of studies, just where you’ve claimed they exist. The authors of this study do cite many papers.

          Here are their conclusions:

          “Conclusions

          Compared with never smokers, current smokers appear to be at reduced risk of SARS‐CoV‐2 infection while former smokers appear to be at increased risk of hospitalisation, increased disease severity and mortality from COVID‐19. However, it is uncertain whether these associations are causal.“

        • You cherry picked Current smoker in this one. The difference with Former smoker is more slight. Other conditions that are less likely to test positive are:
          Cancer: 30% less likely
          Heart failure: 30% less likely

          This is the same as the percentage if you consider “current+former smokers” like you did with the original chinese study.

          One of several things you are missing is that they are comparing people who came into hospital with COVID-19 with people who came into hospital with any other reason. This means that because people go to hospital because they have something wrong with them, the people that come into hospital with COVID-19 are people who tend to be otherwise *healthier* than people who came in because they have some other ailment. This is not a case-control study.

          Incidentally the table also shows that both COVID-19 negative and positive patients are under reporting smoking status. In survey data the % of people who say they are smoking in this age group is 20%.

        • I also remember you trying to draw some kind of conclusion out of asthma having a similar association. Note that asthma in this one has a weaker association than cancer and heart failure.

          Also in case you argue this, it is unwise to use current smoker in this instance because people may terminate smoking as a result of getting COVID-19.

        • Zhou –

          It’s interesting that when you dug through this with a fine tooth comb you found these relevant details.

          Anoneuoid didn’t mention them.

          Did he not find them? Did he find them and “ignore them?”

          Is the eexplanation for that limited to incompetence or dishonesty?

        • Anoneuoid didn’t mention them.

          It is in the quote shown in the reddit thread. They didn’t get the desired results so p-hacked to get what they wanted.

          That’s another thing that makes the missing smokers all the more amazing. The bias against it is huge and still they can’t figure out a way to reliably p-hack it away.

          All the deniers can do is come up with half-baked excuses to ignore it or muddy the waters by conflating different concepts, as we see people continue to do in this thread.

    • I finally got a chance to look closer and that is a nice review.

      I also checked the study reporting the least protection. I saw they included people who had quit in the last five years in their definition of “current smoker”. The next two showing least protection also used the same UK biobank data, but did not clarify the definition of “current smoker”.

      So, hundreds of papers and a year later the conclusion is the same as last spring. For some reason smoking protects against testing positive for SARS viruses.

      • Sorry, I meant since spring 2020.

        Also, that review found smokers are actually ~1.5x more likely to be tested. They still ignore the SARS data from 20 years ago though. That is a key piece of information.

  7. I was surprised that no one referred to the famous
    ——————————————————————-
    https://en.wikipedia.org/wiki/Low_birth-weight_paradox#:~:text=The%20low%20birth-weight%20paradox%20is%20an%20apparently%20paradoxical,It%20is%20an%20example%20of%20Simpson%27s%20paradox%20.
    Low birth-weight paradox – Wikipedia
    The low birth-weight paradox is an apparently paradoxical observation relating to the birth weights and mortality rate of children born to tobacco smoking mothers. Low birth-weight children born to smoking mothers have a lower infant mortality rate than the low birth weight children of non-smokers. It is an example of Simpson’s paradox
    en.wikipedia.org
    —————————————-
    No one on the blog mentioned Simpson’s paradox as the cause/solution of the controversy. The two situations seem so analogous to me so what am I missing?

    • Just couldn’t be bothered. There isn’t much point in arguing with some of the commenters. Too many logical fallacies and other issues and little chance of an impact.

    • No one on the blog mentioned Simpson’s paradox as the cause/solution of the controversy. The two situations seem so analogous to me so what am I missing?

      Like the rest of the alternative theories that have been proposed: you are missing an explanation for why the “missing smokers effect” appears for SARS, disappears for ~20 years, then “coincidentally” reappears for SARS-2…

  8. I don’t think this was linked above, but I thought this is an interesting read in the topic – which includes this interesting note:

    The concept of ‘smoker’s paradox’ is not new. The term was first coined in 1995 to describe the unexpected favourable outcome of decreased short-term mortality in smokers after acute coronary syndrome6 7 and stroke,8 when compared with non-smokers.

    https://ebm.bmj.com/content/early/2020/08/11/bmjebm-2020-111492

    • One of the (to me) interesting aspects not mentioned in that study when they discuss (theoretical) reasons why smoking might be associated with higher rates of covid infection is the question of nocosomal spread. I’d presume that generally, smokers are hospitalized at a higher rate than non-smokers. Given the prevalence of nocosomal spread of covid…

      Other behavioral questions – are smokers apt to spend less time with other people in indoor, poorly ventalated spaces? (Controlling for different national laws limiting indoor smoking might show a relevant signal there.)

      Seems to me, the habit of jumping from cross-sectional data that show correlations to assumptions of causality, is just fundamentally unscientific. Sure, collect those data. They’re important. But then do something with them. Design a study to test different causal mechanisms. In this case, an RCT would be impractical – so then look for other pathways for confirming causality – like doing a deep dive on particular samples to determine whether you see a signal to differentiate former smokers from current smokers, or among current smokers based on rates of smoking. Stratify your data to see if there’s a signal from e-cigarette usage.

      But hey, what do I know?

      • Now you are starting to think about it, good.

        Next consider why the rate in people with asthma is also surprisingly low. Then look into the similarity of covid symptoms with high altitude sickness, and how smoking is also curiously protective against that.

        • Anoneuoid –

          Gathering a list of other correlates, that might be essentially random (or highly conflated, or spurious, or explained by a common mediator), is surely of some use.

          But the point I was stressing is that if you think there might be causality, it seems to me the most logical next thing to do would be to examine that premise rather than assume causality and go fishing for other possible correlates for which you could construct what are essentially arbitrary causal narratives.

          The previous examples of other “smokers paradoxes” should be instructive here.

          But I will acknowledge this is a bit of a pet peeve of mine. I think that a confirmation bias-related tendency towards conflating correlation with caution is a big problem in scientific research. I don’t think there’s actually a “crisis” in research, but to the extent there is one, that’s the “crisis” to me – not so much the use of poor statistical methodology.

          But what I’m more bothered by is that conflating correlation-based analysis with establishing causation is even more of a problem in the scientific research/public discourse interface. There’s been an epidemic of that during this epidemic, imo. Our ability to leverage new technologies to greatly enhance our ability to identify correlates has outpaced our intrinsic habits that govern our tendency towards confirmation bias.

          That Googling has largely become equated to doing research is problematic. Witness Aaron Rogers’ claims of getting “immunized” after doing extensive “research.”

          OK. I should stop before someone calls me an old man yelling at clouds.

        • I will help start you out:

          Fifty-seven studies with an overall sample size of 587 280 were included. The prevalence of asthma among those infected with COVID-19 was 7.46% (95% CI = 6.25–8.67). Non-severe asthma was more common than severe asthma (9.61% vs. 4.13%). Pooled analysis showed a 14% risk ratio reduction in acquiring COVID-19 (95% CI = 0.80–0.94; p < 0.0001) and 13% reduction in hospitalization with COVID-19 (95% CI = 0.77–0.99, p = 0.03) for people with asthma compared with those without. There was no significant difference in the combined risk of requiring admission to ICU and/or receiving mechanical ventilation for people with asthma (RR = 0.87 95% CI = 0.94–1.37; p = 0.19) and risk of death from COVID-19 (RR = 0.87; 95% CI = 0.68–1.10; p = 0.25).

          https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8022341/

          That is surprising right? How would you explain it?

        • I should also quote:

          We found a 14% (95% CI = 0.80–0.94) lower risk of acquiring COVID-19 in people with asthma, which is an absolute reduction of 50 cases per 1 000 people. This is consistent with the trend observed during the SARS pandemic (67).

          Once again, surprising results that show up for both SARS and SARS-2.

        • “Behavioral aspects may have also played a role in reducing the vulnerability of asthmatics to COVID-19 (73). Early in the pandemic, the uncertainty on the impact of asthma on COVID-19 and previous experience of viral infections triggering asthma exacerbations caused anxiety among patients and caregivers (74,75). This followed government advice during the peak of the pandemic in countries like the United Kingdom, which classified severe asthmatics as a vulnerable group and advised them to shield for 12 weeks at home (5). A study in USA showed that during the pandemic, there was a 14.5% relative increase in daily controller adherence in asthmatics and COPD patients which supports this posit (76). All these factors may have worked together in reducing the risk of acquiring COVID-19 in people with asthma.”

        • Yep, you can keep coming up with different explanations for why an apple falls to the ground and the moon “coincidentally” also orbits the earth. For all practical purposes, there is an infinite combination of these ideas to squander your time and resources on.

          Or you can do science and try to find a common explanation that leads to it all making sense. Strive for that, then you get progress and previously unimagined benefits. I say use science.

        • You don’t understand basic science, by your own admission. You made a laughably poor prediction regarding weekly COVID cases–a prediction that should have been obvious (to you, at the time) because anyone that understood ‘basic science’ would have made the same prediction. You also conveniently ignored other people that were willing to take you up on your bet. It would have been nice for you to go broke from your own hubris and self-inflated ego.

          I am not “coming up with different explanations for why an apple falls to the ground.” The quote above was from the same meta-analysis you are using to push your foolish narrative. Why don’t you read what you link, troll!

        • > The quote above was from the same meta-analysis you are using to push your foolish narrative.

          Yes, of course. Where did I say otherwise?

          You may not be a native english speaker. I was using the “royal you.” In general, “you” (or “we”) have a choice between viewing the world as a collection of disconnected events, or use science to figure out how these events may be connected to each other according to a common explanation. The case of smoking and asthma vs covid is obviously one of those times.

          > You made a laughably poor prediction regarding weekly COVID cases

          Actually, I predicted they would trough at the end of Oct. There is a post here from before the bet thread explaining I was “giving a few extra weeks” to entice more attention on the bet.

          But even using the Oct 6th date, the results were not “laughably poor”.

  9. Ah, a Farsalinos study! Someone just pointed me to this discussion today, so apparently? people still take them seriously…

    0. Does smoking have a protective effect? Seems doubtful. Here’s an N=420,000 study: https://thorax.bmj.com/content/77/1/65

    1. Most smaller studies (on smoking and SARS) fail to overcome methodological challenges. Hospitalizations that do happen may be a bit more severe; maybe there is a ‘missing cohort’ of mild cases that don’t get hospitalized or detected. There have been zero studies that are compelling enough for anyone to try replicating them, warts and all. The last Frontiers paper Anon posted simply shows that among the patients with “any respiratory symptom, fever, risk factors for COVID-19 complications” or recent exposure, which would lead to a PCR test, smokers were less likely to test positive.

    1.9 Nicotine research to identify new markets for the industry continues apace, and has done since well before 2003.
    https://tobaccocontrol.bmj.com/content/28/3/289.abstract

    2. Is the original paper crap, rather than “something worth considering”? Almost certainly.
    Farsalinos published this as part of a series of preprints (most with K. Poulas) in March-May 2020 : that inspired a series of follow-ups by others. They ignored field norms by publishing basically the same paper in a range of journals across discipline. They were(?) smoking-industry funded – as in “run a commercial pro-nicotine-research institute”, not just “got funding from a tobacconist once”. Failure to disclose led to one of this family of papers being retracted.
    https://tobaccotactics.org/wiki/nosmoke-institute/

    2.1 F continued to push this line, throughout the year. “Could nicotine be a therapeutic option for COVID-19?” https://pubmed.ncbi.nlm.nih.gov/32385628/
    Despite this, post-retraction he tried to argue that there was no conflict of interest. https://www.qeios.com/read/99DA06

    2.2 F later became popular among a circle of anti-COVID-vax coauthors and pundits, and publicly supported research claiming vaccines kill more than they save. Neither here nor there, but another reason not to trust his data w/o replication.

    3. I don’t believe there is a “known link” between smoking and protection from either SARS or SARS-2. Contrary to the claim above that “this is a known correlation” “in every country” which “only affects SARS and SARS-2”, I’ve seen no strong result or replication, occasional anti-replication (see 0.) , and enough confounds and bad studies to keep a nicometa site happy. There /is/ a popular myth about this; popular enough to drive some small studies. The NPJ paper mentioned gives a fine summary. https://www.nature.com/articles/s41533-021-00223-1

    *Could* there be some low-effect-size protective effect? Certainly. Is that likely, given research to date? Signs point to no.

    • You are just not informed and are confusing different issues (getting covid vs getting severe covid given you already got covid). These were the final results of the largest meta-analysis:

      Conclusions: Compared with never smokers, current smokers appear to be at reduced risk of SARS-CoV-2 infection and increased risk of greater in-hospital disease severity, while former smokers appear to be at increased risk of hospitalisation, greater in-hospital disease severity and mortality from COVID-19. However, it is uncertain whether these associations are causal. This version (v12) will be the last regular update; however, yearly updates may continue as new evidence becomes available.

      https://www.qeios.com/read/ujr2aw.15

      And yes, they ignore the most informative data. Which is that this shows up for SARS and is obvious to everyone but statisticians, then disappears for nearly 20 years, and reappears for SARS-2. That is the convincing piece of evidence that something is going on here.

  10. We identified female sex, airborne allergy, and smoking to be associated with protection from Covid-19, which is in line with previous findings. A Spanish registry study covering close to half a million individuals showed that the risk of contracting Covid-19 was lower in asthmatics with an odds ratio of 0.74 (95% CI: 0.71–0.77) [24]. Although it is clear that smoking is a risk factor for the severity of Covid-19, early studies reported an underrepresentation of smokers among patients hospitalized for Covid-19 [25]. Perhaps smoking-induced inflammation of the upper respiratory mucosa provides low-degree protection against transmission of viral infection.

    https://onlinelibrary.wiley.com/doi/10.1002/eji.202149655

  11. During the study, 44,270 patients had SARS-CoV-2 infection. Current smoking was associated with lower adjusted rates of SARS-CoV-2 infection (aHR=0.64 95%CI:0.61-0.67), COVID-19-related hospitalization (aHR=0.48 95%CI:0.40-0.58), ICU admission (aHR=0.62 95%CI:0.42-0.87), and death (aHR=0.52 95%CI:0.27-0.89) than never-smoking. Former smoking was associated with a lower adjusted rate of SARS-CoV-2 infection (aHR=0.96 95%CI:0.94-0.99) and higher rates of hospitalization (aHR=1.10 95%CI:1.03-1.08) and death (aHR=1.32 95%CI:1.11-1.56) than never-smoking. Logistic regression analyses among patients with COVID-19 found lower odds of hospitalization for current versus never-smoking and higher odds of hospitalization and death for former versus never-smoking.

    https://academic.oup.com/ntr/advance-article-abstract/doi/10.1093/ntr/ntac090/6562890

    • > current cigarette smoking was associated with a lower risk […] severe COVID-19 illness compared to never-smoking. A history of smoking was associated with a […] modestly higher risk of severe COVID-19 illness compared to never-smoking.

      Well that’s an interesting finding. Not able to read the paper, I’m left wondering what they think might explain it.

      And of course, the paper makes it clear there’s no controls for behavior. I don’t get how anyone could think such a finding is robust without any such controls.

      • I don’t get how anyone could think such a finding is robust without any such controls.

        I have said since the beginning that the convincing aspect is it shows up for SARS, disappears for 20 years, then reappears for SARS-2. Also that asthmatics seem protected as well. This implies that chronic inflammation is doing something protective, perhaps more ACE2 on goblet cells found in the upper airways.

        In general, I don’t think the “adjusted” coefficients that come from these models that incorporate whatever data is lying around mean much. In this case it is a bit more convincing since there is such a strong bias against seeing this effect though. If anything people will tend to p-hack to get rid of the effect.

Leave a Reply

Your email address will not be published. Required fields are marked *