Last Orders with Mark Littlewood

I had a blast recording a new Last Orders episode on Friday with Tom Slater and Mark Littlewood. It's the first time the show has been recorded in person for about eighteen months.

We discussed the Extinction Rebellion cult, smoking and Zero Covid. Listen here.

A Swift Half with Stuart Ritchie

The latest Swift Half with Snowdon features Dr Stuart Ritchie, a psychologist at KCL and the author of Science Fictions. We discuss IQ, fraud in psychology and Covid 'sceptics'. 

The WHO's war on alcohol

Last year the World Health Organization launched a public consultation on its draft “Global Alcohol Strategy to Reduce the Harmful Use of Alcohol”. It is not obvious that the world needs an alcohol strategy, nor does it seem an obvious priority for the WHO in the middle of a pandemic. National governments are quite capable of deciding how alcoholic drinks are taxed and regulated without pressure from a UN agency. Some countries allow you to buy a beer at any time day or night. Others have total prohibition. The huge differences in the way in which governments treat alcohol make it an unlikely candidate for global regulation, but the World Health Organization is keen to leave its mark on the issue nonetheless. 

The WHO has had a Global Alcohol Strategy since 2010, but the draft of the new strategy represented a significant shift in emphasis. While the 2010 plan acknowledged the problems created by the illicit market, which makes up 25% of global alcohol supply and more than 50% in some countries, the working document barely mentioned it. And while the 2010 focused on alcohol-related harm, the working document treats mere consumption of alcohol as a problem in itself.

This went largely unreported in the media because the WHO gave journalists a better story when they said that more efforts should be made to prevent “women of childbearing age” from drinking. This was yet another public relations disaster from an agency that has careered between tragedy and farce under the leadership of Dr Tedros Adhanom Ghebreyesus. From making Robert Mugabe a “goodwill ambassador” in 2017 to allowing Covid-19 to spread rather than offend the Chinese, the WHO has had a bad few years.

After launching a public consultation on its alcohol plan, the WHO came back with a new draft last month in which the reference to childbearing women was wisely deleted. Elsewhere, however, the problems remain. The new draft is actually even worse because the authors have inserted an unjustifiable and wholly unrealistic target of reducing per capita alcohol consumption worldwide by 20 per cent by 2030.

There was no target for per capita alcohol consumption in the 2010 plan. There was a target in the WHO’s Global Non-Communicable Disease Action Plan of 2013 but that was to reduce the “harmful use of alcohol, as appropriate, within the national context” by 10 per cent. The new proposal of a 20 per cent reduction in per capita alcohol consumption cannot be justified on health grounds since consumption is not a measure of health, and it is patently unachievable in such a short space of time. There is no reason to believe that this target could be met even if every member state introduced the WHO’s so-called “best buys” (tax rises, advertising bans, etc.) tomorrow.

Where did this target come from? A clue may lie in last year’s consultation which was inundated with submissions from temperance organisations. Many of these submissions were very similar. Some were identical. One of the biggest organisations in this network is Movendi International which was, until recently, known as the International Order of Good Templars. It was formed in the 1850s to campaign for the total prohibition of alcoholic beverages, but has since taken a more incremental approach to building an alcohol-free world. In its submission, Movendi called on the WHO to set a target of a 30 per cent reduction in per capita consumption by 2030, saying: “We propose a bold and ambitious overall target of a 30% reduction of per capita alcohol consumption until 2030”.

This exact phrase was repeated word for word in the consultation responses of numerous other organisations, ranging from IOGT Iceland and Slovenia’s Institute for Research and Development to Cambodia’s Khmer Youth Association, Kenya’s Alcohol Control Policy Network and Tanzania’s Network Against Alcohol Abuse. Duplicate responses were particularly common from temperance and public health organisations in Asia and Africa.

Organisations which endorse total abstinence from alcohol will naturally support any measures to suppress consumption. They can be expected to propose the most extreme targets for reductions in alcohol consumption, regardless of how unrealistic they may be. But pressure groups rooted in the temperance movement have very little public support and it is concerning to see the preferential treatment given to them by the WHO. Movendi, for example, is listed as a “non-state actor in official relations” with the WHO. Its spokespeople regularly appear at WHO conferences and it has an official arrangement with the WHO to “draft advocacy materials linked to WHO’s activities”, “increase support for alcohol control initiatives” and to produce “technical sessions and webinars to discuss and promote WHO’s global public goods in the field of alcohol prevention and control with content reviewed and approved by WHO in line with its policies and guidelines”. Movendi has also previously agreed to “help explore new or innovative ways and means to secure adequate funding for the implementation of the WHO Global strategy to reduce the harmful use of alcohol”.

In light of the close relationship between the WHO and the global temperance movement, it is reasonable to ask whether the WHO has simply split the difference between the 30 per cent target for consumption proposed by groups such as Movendi and the 10 per cent target for harmful use already in place in the NCD Action Plan to arrive at a target of 20 per cent for consumption. It is difficult to see any other logic to the decision since there is no evidence that a 20 per cent reduction is attainable or optimal. It is a false compromise; the “golden mean fallacy” writ large.

Since there is no prospect of this target being met, the WHO is setting up member states to fail. When 2030 comes around and this arbitrary and unrealistic milestone has not been reached, we can expect the WHO and its temperance partners to demand even tougher action, perhaps including a legally binding Framework Convention on Alcohol Control (modelled on the Framework Convention on Tobacco Control) that has so far been firmly rejected by member states.

Whatever the reasons for the sudden insertion of this target into the Global Action Plan, it is quite unacceptable. Insofar as the WHO has a mandate for getting involved in alcohol policy, it is in relation to health harms, not consumption. There is no reason to assume that a reduction in per capita consumption will necessarily lead to a reduction in alcohol-related harm. As the latest draft of the Plan acknowledges, wealthier countries tend to have higher rates of consumption but do not have higher rates of heavy episodic drinking.

The focus on consumption allows the WHO to push ahead with the kind of crude, supply-side policies that are popular with western public health academics but which can only discourage the sale of legal alcohol. Meanwhile, they virtually ignore illicit alcohol which is a bigger problem in most low and middle income countries. In contrast to the WHO’s 2010 plan, the new draft says almost nothing about the social, economic and health harms of black market booze and does not acknowledge the risks of increasing demand for homemade and illicit products by suppressing demand for legal products.

A new public consultation has now been launched and will run until 3 September. No doubt the temperance lobby will complain that the draft is still not extreme enough, but the WHO should listen to more moderate voices before it embarks on a plan that is designed to fail and which will be wholly unacceptable to many member states.

Cross-posted from the IEA blog. See also Henry Hill's article about the WHO at CapX.

A critique of the sugar panic

I've been reading an interesting critique of the current hysteria about sugar published in Critical Reviews in Food Science and Nutrition. Here are some highlights...

The pursuit of power and profit via the demonization of dietary factors such as alcohol and sugar dates to at least the seventeenth century. Thus, as the public’s fear of fat and cholesterol waned and the lipid-heart disease hypothesis began to lose empirical support, it was only a matter of time before media attention, political grandstanding, and research funding were once again directed at dietary sugars. The net result of this shift in attention and resources was an apparent ‘consensus’ that sugar consumption is harmful.

.. because a million matching but ill-informed opinions do not constitute a fact, the anti-sugar consensus, either real or apparent, is not a statement about nutrition science but merely the ‘status quo’ made explicit.

.. the distinctions between ‘added’, ‘free’, and ‘intrinsic’ sugars are scientifically and biochemically mean- ingless. As such, the use of the term ‘added sugar’ in a scientific or health context is misleading. Nevertheless, the ubiquitous use of the term in popular and social media led to ‘added sugars’ becoming “the nutrition villain du jour” (Slavin 2014, p. 4) with subsequent ill-informed changes in the public’s perception of sugar-health relations

Importantly, by recommending breast-feeding, public health organizations are implicitly arguing that infants should consume $40% of their total calories as dietary sugars. Yet these same organizations also recommend that sugar consumption by infants and children be limited. The striking contradiction between the passionate prescription of breast-feeding and the puritanical proscription of dietary sugars is due to the failure to acknowledge that, despite its political expediency and marketing potential, the distinction between ‘added’ and ‘intrinsic’ sugars is bio- chemically and scientifically meaningless.

In the U.S., sugar availability (i.e., a crude proxy for consumption) increased $600% from 18.6 lbs (8.5kg) in 1863 to 125 lbs (57kg) in 1930. The fastest rate of increase in U.S. history was 3 lbs per capita per year (36.1%) from 1918 until 1930. Yet as sugar availability increased exponentially, the U.S. population improved in almost every health metrics (e.g., malnutrition, mortality, health-span) while rates of obesity and T2DM were extremely low.

Diet-centric paradoxes exist because investigators often ignore contrary evidence or refuse to question assumptions that have been demonstrated to be false. For example, there are countries in which sugar and sugar- sweetened beverage consumption declined while obesity and metabolic diseases increased. In Australia, sugar/sweetener availability decreased 18% from 56.2 kg in 1961 to 46.1 kg in 2013 as the prevalence of obesity increased over 450% from more than 5% in the 1960s to 28% in 2015). In the United Kingdom (U.K.) availability decreased 20% from 51.7kg in 1961 to 41.3kg in 2016 while obesity increased over 400% from less than 5% in the 1960s to over 26% in 2016. In the U.S., sugar/sweetener availability declined 18.2% from 69.1 kg per capita in 1999 to 56.5 kg in 2018, as the prevalence of diabetes in adults increased 36.8%, obesity increased 39%; and severe obesity increased 95.7%

The end result of the intellectual decline and corruption in nutrition research has been decades of impassioned but physiologically illiterate disputes about sugar, salt, fat, and cholesterol, increasing confusion about what constitutes a ‘healthy diet’, and the public’s loss of confidence in ‘science’ to inform their lives. 

.. physiology, not food and beverages, causes metabolic diseases. This fact explains why identical diets consumed by different individuals result in divergent nutritional, metabolic, and health effects, and why some individuals can consume massive quantities of sugar and other carbohydrates while maintaining metabolic health, whereas less fortunate individuals develop obesity and/or T2DM. 

We contend that current policy recommendations on ‘added’ sugars and sugar-sweetened beverages are not only unscientific, but also regressive and unjust because they harm the most vulnerable members of our society while providing no personal or public health benefits 

 

It's paywalled but there are ways and means of downloading it.

Signs of sanity in US tobacco control?

The American Journal of Public Health has published an article by fifteen past presidents of the Society for Research on Nicotine and Tobacco. Not all of it is sensible - they support raising the age at which e-cigarettes can be bought to 21 and they support the idiotic policy of reducing nicotine in cigarettes - but it is a beacon of enlightenment compared to most of the commentary coming out of the USA these days.

The key points:

Vaping reduces the smoking rate

For years, US cigarette sales declined 2% to 3% annually. More recently, as vaping product sales increased, cigarette sales decreased much more rapidly. Con- versely, following the EVALI outbreak and e-cigarette sales restrictions, sales of e-cigarettes fell and sales of cigarettes resumed their prevaping pattern. Studies finding a positive cross-price elasticity of demand between cigarettes and e-cigarettes support the conclusion that the products are substitutes. 

Although not the final word, the totality of the evidence indicates that frequent vaping increases adult smoking cessation. Smokers unable to quit smoking with evidence-based cessation methods should be well informed about the relative risks of vaping and smoking and vaping’s potential to help them quit smoking.

Flavour bans are stupid an counterproductive

Recent policy attention has focused on restricting the availability of e-cigarettes with flavors, a principal attraction for youths. While flavor bans could reduce youth interest in e-cigarettes, they could also reduce adult smokers’ vaping to quit smoking. Like youths, adults prefer nontobacco flavors, both groups favoring fruit and sweet flavors. Policies regarding flavors reflect the more general issue considered in this article: the need to create a balance between the sometimes-conflicting goals of preventing youth vaping and supporting adults’ smoking cessation attempts 

The gateway effect is trivial, if it exists at all

Prospective studies have found that young people who had vaped but never smoked cigarettes were more likely to have tried cigarettes several months to 2 years later than contemporaries who had neither smoked nor vaped. Some commentators thus consider vaping a “gateway” into smoking. Other observers believe the relationship reflects a “common liability”: young people who vape are generally more prone to risky behavior; hence, they might be more likely to try smoking even without vaping. Three recent studies have concluded that vaping likely diverts more young people from smoking than encourages them to smoke.

.. If vaping causes some young people to try cigarettes, the aggregate impact must be small. 

..Vaping may addict some youths to nicotine, but many fewer than popularly believed. 

Anti-vaping policies are effectively pro-smoking policies

To date, the singular focus of US policies on decreasing youth vaping may well have reduced vaping’s potential contribution to reducing adult smoking. Those policies include taxing e-cigarettes at rates comparable to cigarette taxes, decreasing adult access to flavored e-cigarettes that may facilitate smoking cessation, and convincing the public—including smokers—that vaping is as dangerous as smoking.

Conclusion

We share the very legitimate concerns about youth vaping with the entire field of public health. Our goal is to put those concerns in perspective. We agree with former Surgeon General C. Everett Koop who, in 1998, urged that “[A]s we take every action to save our children from the ravages of tobacco, we should dem- onstrate that our commitment to those who are already addicted . . . will never expire.” The latter appears at risk today.

While evidence suggests that vaping is currently increasing smoking cessation, the impact could be much larger if the public health community paid serious attention to vaping’s potential to help adult smokers, smokers received accurate information about the relative risks of vaping and smoking, and policies were designed with the potential effects on smokers in mind. That is not happening.

Do read it all.

Vaccines, vaping and varenicline

In my City AM column today, I discuss the risks of vaccines, vaping and varenicline and the contradictory ways in which they are treated by regulators and the WHO.

When the AstraZeneca vaccine was found to be associated with blood clots in a small number of cases, the World Health Organisations’s Global Advisory Committee on Vaccine Safety urged its continued use because, it said, “the AstraZeneca vaccine’s benefits outweigh its risks”. It added that: “The question with any pharmaceutical or vaccine is whether the risk of taking it is greater or less than the risk of the disease it is meant to prevent or treat.”

And yet the WHO is also fiercely opposed to e-cigarettes. A WHO report published last month called for countries to either ban their sale completely or subject them to harsh, tobacco-style regulation. It is eager to stress that “e-cigarettes are harmful to health” although it is far from clear what those risks are. When it comes to vaccines, the WHO is interested in relative risks and the overall effect on the health of the population. When it comes to vaping, however, it cannot see beyond tiny – and largely hypothetical – absolute risks.

Do read it.

Randomised controlled trial shows that plain packaging doesn't work

Nearly a decade after plain packaging was introduced in Australia, a randomised controlled trial has been conducted to see if the policy actually works. The trial was registered last September and its results were published last week.

You probably missed it. It received virtually no media coverage, for reasons that will soon become obvious, and it was confusingly portrayed as being a study of graphic warnings. Indeed, the study is titled 'Effect of Graphic Warning Labels on Cigarette Packs on US Smokers’ Cognitions and Smoking Behavior After 3 Months'.

The researchers gathered a group of 357 smokers and split them into three groups. Participants were sold cigarettes for four months (online delivery). In the first month they were sold cigarettes in a regular, US pack with no graphic warnings and a relatively discreet written warning. Thereafter they were sold cigarettes either in a regular US pack, a totally blank pack or a pack with graphic warnings on them. 

It is only at the end of the study that it becomes clear that the pack with graphic warnings is actually what we know as a plain pack. In fact, they literally used plain packs from Australia.

In this study, we chose the Australian standardized packaging for our GWL packs rather than the hybrid packs proposed for implementation in the US.

These are shown in one of the appendices:

During the course of the study, the researchers sent participants two text messages a day asking if they agreed with various statements (such as "my last cigarette was satisfying") and asking how many cigarettes they had smoked in the last four hours. They also sent them a weekly survey asking questions such as "How often did you think about wanting to quit?"

The participants also took part in two detailed surveys during the course of the study and were tested for cotinine to confirm their smoking status.

So what happened? In line with several other studies, smokers exposed to the graphic warnings had a slightly less positive perception of cigarettes. They also expressed a "marginal increase" in health concerns and declared themselves to be slightly more interested in quitting.

But this didn't lead to any behavioural changes. They didn't quit. They didn't reduce the number of cigarettes they smoked. They weren't more likely to manage even four hours without a cigarette.  

Across the first 2 months of the intervention, only those in the GWL [graphic warning labels] pack group reported a consistent decrease in their positive perceptions of the cigarettes that they smoked, although the expected increase in perceived health concerns was not significant. Although there was a significant increase in cognitions about quitting in the GWL group, there was no evidence of increased quitting or reduced consumption, and this was biochemically validated at the postintervention visit.

This seems to have come as a surprise to the authors.

Health concerns increased in all study groups during the intervention, with only a marginal increase in the GWL pack group. As the theory supporting mandated GWLs expects that their effect will be achieved through increased health concerns, this lack of a marked between-group difference was unexpected. 

.. Although the GWL group participants were more likely to think about quitting during the study intervention, there was no evidence of increased quitting behavior. Our measure of at least 1 weekly 4-hour smoking abstinence was designed to maximize the possibility of identifying early quitting activity. What was surprising is that the GWL group participants, who were more likely to think about quitting, were not much more likely to report these short smoking abstinences or to report lower daily cigarette consumption. 

If they were surprised, they shouldn't have been. Because...

Our study is in line with previous research indicating that intentions to change are rarely sufficient to achieve change in an addictive behavior.

So why don't graphic warnings work?

That the GWL pack did not experience a significant increase these cognitions may reflect the high level of awareness of the health consequences of smoking

Indeed. Smokers are well aware of the risks, as are nonsmokers who take up the habit. If anything, they have an exaggerated perception of the risks. We have had evidence for over a decade telling us that hammering home the risks with graphic warnings makes no difference to smoking rates regardless of what smokers might tell researchers in focus groups.   

The campaigns for both graphic warnings and plain packaging were driven by studies showing that people expressed certain views when confronted with the new packs. Often these boiled down to little more than them saying that they found an ugly pack ugly and an attractive pack attractive. At other times, they expressed a vague desire to stop smoking. 

But talk is cheap and subjective evidence in the field of tobacco control is riddled with social desirability bias. Stated preferences are worthless if they are not accompanied by action. This study was explicitly "designed to maximize the possibility of identifying early quitting activity". The researchers would have settled for four hours abstinence, but the participants couldn't even manage that.

The authors of the study, who are based in California, attempt to salvage something from it by talking up the impact of graphic warnings on "positive perceptions of cigarettes" and by speculating that these packs could be effective if combined with other (unspecified) policies, but the results speak for themselves.

There is a huge difference between a standard American cigarette pack with a modest health warning and a standardised Australian cigarette pack with no branding whatsoever and a large graphic warning. And yet abruptly switching from one to the other made essentially no difference to smokers' behaviour.

This randomised controlled trial shows that graphic warnings are essentially useless, but it also shows that plain packaging doesn't work. Two birds, one stone. The use of a totally blank cigarette pack as a third option in the study helps to underline how futile the branding ban aspect of plain packaging is. Removing the branding didn't even change perceptions, let alone behaviour.

Tobacco branding on a cigarette pack typically increases a smoker’s perceptions of satisfaction, craving relief, and taste of cigarettes. However, in this study, removal of tobacco branding on the blank pack was not sufficient to decrease these positive perceptions.

It seems a shame that nobody bothered to conduct research like this before governments around the world committed themselves to a policy that doesn't work. But actually they did. A smaller RCT conducted in Britain in 2015 came to much the same conclusion: smokers don't particularly like plain packaging but it doesn't have any effect on their smoking habits. 

The British government went ahead with the policy anyway.

Alcohol-related deaths rose sharply in Scotland last year

The alcohol mortality figures for Scotland were published today and 2020 saw a significant increase. This was expected because it is what we saw in England and Wales. In England and Wales, the death rate rose last year by 18% to 13 per 100,000 people. In Scotland, it rose by 16% to 21.5 per 100,000 people.  

Interestingly, in Scotland the increase was almost entirely due to more men dying of alcohol-specific causes whereas in England and Wales both sexes were affected more or less equally. 

People will probably argue for years about the exact mechanisms behind this. The bottom line is that more people drank themselves to death last year than in a normal year and this is probably due to fear, despair and boredom, partly caused by the pandemic itself and partly caused by the social isolation caused by stay-at-home orders and the closure of social venues. 

There is evidence that heavy drinkers consumed more alcohol under lockdown while lighter drinkers consumed less. There is also evidence that overall alcohol consumption declined under lockdown in England but stayed the same in Scotland. In neither country was there an increase in alcohol consumption, despite 'public health' dogma suggesting that an increase in consumption is a prerequisite for an increase in harm.

Due to pubs, clubs and restaurants being closed for much of the year, alcohol was less 'available' in 2020 than usual. It was less advertised in 2020 for the same reason. Whether it was less affordable is a moot point. On average, it was cheaper because alcohol in the on-trade is more expensive, but the pandemic did not make off-trade alcohol any cheaper than usual. 

In Scotland, off-trade alcohol is more expensive than in England thanks to minimum pricing, but this did not prevent Scotland having a similar spike in deaths as seen in the rest of Britain. 

The three factors temperance campaigners claim are the key drivers of alcohol harm - availability, affordability and advertising - either declined or stayed the same in 2020 and yet alcohol harm - as measured by deaths - rose sharply.

What changed were people's circumstances. 2020 was a horrific year characterised by stress, loneliness and tedium. It is no surprise that alcohol-related mortality rose. Drug-related deaths also rose, particularly in Scotland. The success or failure of minimum pricing cannot be judged by what happened in this exceptional year, but the similarity between the mortality spikes in Scotland and England, despite minimum pricing being in place for two years, shows that such policies are, at best, trivial in the context of an individual's personal circumstances. 

If you want to deal with the deaths of despair, start by preventing despair. Everything else is tinkering, and much of it creates its own problems.

PS. Inevitably, the Alcohol Health Alliance has responded to today's news by calling for advertising bans and higher alcohol taxes. 

The drinking guidelines were set by temperance zealots

First published by Spectator Health in February 2016 


I've been reading the minutes of the meetings held by the committee that reviewed the alcohol guidelines recently. You may recall that this was the first full review since 1995 and led to the Chief Medical Officer (CMO), Sally Davies, lowering the recommendations for men from 21 to 14 units a week. The female guidelines were left at 14 units. She also claimed that the health benefits derived from moderate drinking were an ‘old wives’ tale’ and claimed that there was ‘no safe level of alcohol’.

The most striking difference between the 1995 review and the 2016 review is the make-up of the panels. Whereas the 1995 committee was dominated by civil servants who had no obvious prejudices for or against alcohol, the meetings held from March 2013 to discuss alcohol guidance were dominated by activist academics and temperance campaigners.

The Institute of Alcohol Studies (IAS), a small but hardline anti-alcohol organisation, was heavily represented on the committee. The IAS was formed in the 1980s as a direct successor to the UK Temperance Alliance which, in turn, had been formed out of the ashes of the UK Alliance for the Suppression of the Traffic of All Intoxicating Liquors, a prohibitionist pressure group. The IAS receives 99 per cent of its income from the Alliance House Foundation whose official charitable objective is ‘to spread the principles of total abstinence from alcoholic drinks’. Its director, Katherine Brown, was on the CMO’s panel, as was its ‘expert adviser’ Gerard Hastings, although he failed to disclose his IAS role in his declaration of interests.

The IAS’s scientific adviser Petra Meier was also on the committee and was joined by her Sheffield University colleague John Holmes. Holmes and Meier are both strong advocates for minimum pricing and helped develop a computer model which has been repeatedly used to promote minimum pricing by producing estimates of the number of lives that will supposedly be saved by the policy.

Another staunch anti-alcohol campaigner, Ian Gilmore (chairman of the Alcohol Health Alliance, of which the IAS is a key member), was unable to attend the first meeting but was involved thereafter. Gilmore has campaigned for many years for higher alcohol taxes, minimum pricing and a total ban on alcohol advertising.

Other members of the committee may have been less strident than Gilmore and the IAS but there was no doubt where their biases lay. Mark Bellis wrote an article for the British Medical Journal in 2011 complaining that existing alcohol guidance was too generous and ‘read more like an alcohol promotion slogan’. Mark Petticrew and Theresa Marteau are both strong advocates of a range of heavily interventionist ‘public health’ policies, including sugar taxes, plain packaging and minimum pricing.

Of those who attended the initial meetings in 2013, only three did not explicitly advocate stricter alcohol control: the health economist Martin Buxton, the health sociologist Sally Macintyre, and the epidemiologist Valerie Beral. The latter appears to have been selected because of her research linking alcohol to breast cancer, which would become a crucial element in the ‘no safe level’ narrative.

The minutes of a meeting in June 2013 indicate that the path towards dismissing the benefits of moderate alcohol consumption was mapped out from an early stage. Mark Petticrew told his new colleagues: ‘The beneficial effects of alcohol consumption, where they are evidenced, are limited to a low consumption level of half a drink per day.’ Moreover, he said, ‘The population cohort who experiences any beneficial health effect from alcohol is very small. Given these limitations, there is an argument that beneficial effects could be considered not to be relevant in the context of an overall population message, advice or guidance.’

These bald assertions were based on private meetings held between Petticrew and two alcohol researchers, Jurgen Rehm and Tim Stockwell, a fortnight earlier. As a result of this information, the minutes of the 25 June meeting state that the group ‘agreed that a key message from the Rehm/Stockwell discussion is that the evidence shows that any amount of alcohol increases the risk of cancer. Therefore it cannot be said that there is such thing as a “safe” limit.’

Here were the two central messages that would be transmitted, almost word for word, to the British public two and half years later — that the benefits of moderate consumption had been much exaggerated and there is no safe level of drinking. This new narrative appears to have arisen from nothing more than a private meeting with two researchers. Rehm has strong views on alcohol policy (he advises governments to ‘treat alcohol like tobacco’) and he is a respected alcohol researcher but his views, as ventriloquised by Petticrew, bear little relationship to what he told the BBC after the guidelines were announced. On BBC Radio 4’s More or Less programme, he made it clear that there was good evidence that moderate alcohol consumption reduced the risk of heart disease and other diseases. Rehm’s own research concluded that the protective effect of alcohol on heart disease was ‘hard to deny’ and not just for those who consume ‘half a drink a day’, as Petticrew claimed, but for larger quantities too. Rehm’s 2012 systematic review found that heart disease risk was at its lowest for men drinking around four units a day, with a lower optimal level for women.

While Rehm’s views may have been misreported by Petticrew, those of Tim Stockwell were not. Stockwell is the world’s most persistent and prominent critic of the evidence showing that moderate alcohol consumption saves lives. His various letters, editorials and studies casting doubt on the benefits of drinking were given a hugely disproportionate prominence in the 2016 guidance. It is telling that Petticrew’s first act was to approach Stockwell and allow his controversial opinion to frame the debate.

The other striking difference between the 1995 review and the 2016 review is the range of evidence put before the respective committees. Whereas the 1995 panel received dozens of submissions, the minutes of a March 2013 meeting show Sally Davies’s team explicitly rejecting a call for evidence, preferring instead to rely on their own wisdom. Several new reports were commissioned, but all were co-authored by members of the committee.

Most of these were commissioned from the Centre for Public Health at Liverpool John Moores University and co-written by Mark Bellis. One of them, entitled ‘A summary of the evidence of the health and social impacts of alcohol consumption’, did its utmost to cast doubt on the benefits of alcohol consumption.

Sceptics such as Stockwell often claim that non-drinkers have a lower life expectancy than moderate drinkers because many of them are unhealthy former drinkers. Despite many studies showing that moderate drinkers also live longer than lifetime abstainers, this zombie argument continues to be made and it reappeared in Bellis’s report for the committee. He was, however, forced to admit that studies which have controlled for this potential confounder still found a protective effect. The draft document concedes: ‘A few meta-analyses have sought to account for such bias, and based on the extent to which this misclassification error can be accounted for, compared with lifetime abstainers a protective association appears to remain for type-2 diabetes, ischaemic heart disease, and ischaemic stroke.’

This was undoubtedly true. The Liverpool report included a summary of epidemiological studies showing that risk from several major diseases is lowest for people drinking between 1.5 and 8.5 units a day and that risk only reverts to that of an abstainer at a level of at least 4 units a day (or 28 units a week). The authors did not dwell on this evidence. Instead, they immediately suggested that there were ‘further reasons to suggest that the beneficial effects of alcohol consumption may currently be overestimated’, a vague claim for which the only citation was an opinion piece by Tim Stockwell.

The Liverpool document has since been made available to the public but the published version has been edited to further obfuscate the benefits of drinking. Whereas it previously acknowledged the evidence that moderate drinking reduces the risk of type-2 diabetes, ischaemic stroke and heart disease, it now only mentions heart disease and Stockwell’s opinion is given added prominence. The passage quoted above has been replaced by the following: ‘A few meta-analyses have sought to account for such bias; for example a recent meta-analysis, which reported that light to moderate alcohol consumption was associated with a reduced risk of cardiovascular outcomes, included lifetime abstainers as a reference category in sensitivity analyses. However, Stockwell et al question the robustness of the conclusions generated from this literature…’

By November 2013, Mark Petticrew had already drafted the committee’s conclusions. He acknowledged that many studies have found a ‘J-shaped relationship between alcohol consumption and total mortality’ but after considering evidence from Liverpool John Moores University and his conversations with Rehm and Stockwell (no other evidence was mentioned) he claimed that the ‘estimates of the size of this protective [effect] are likely to be biased’. Petticrew had no such concerns about flaws in the epidemiology of cancer, however: ‘For cancers there is clear and consistent evidence of a linear relationship. Alcohol is carcinogenic with no safe lower limit.’

The draft guidelines concluded that the benefits of alcohol consumption, such as they were, mainly affected people over the age of 50 and only related to heart disease. The latter is untrue (the evidence before the committee clearly showed a protective effect for other diseases) and the former is largely irrelevant (heart disease is rare among people under 50). Despite portraying the benefits as only applying to older people, Petticrew advised against telling them to drink alcohol. ‘Discussion at previous meetings,’ he wrote, ‘was along the lines of: if someone >65 is not currently drinking, then the evidence is not strong enough to recommend them to start; however if they are currently drinking more than the lower limit, then they should reduce their consumption.’

It was becoming clear that the bar for what constituted good evidence was being set much higher for benefits than it was for risks. The idea that the government should recommend moderate alcohol consumption to people who did not drink was regarded as unthinkable, regardless of the health benefits.

When a second draft of the guidelines was written at the end of January 2014, an even harder line was taken and a new argument had been found. Having whittled away the benefits of drinking until they applied only at a low level to a single disease among one section of the population, Petticrew explained that heart disease in Britain was not the killer it once was and, therefore, ‘irrespective of whether any protective effect is real or artefactual, any positive impact on total mortality is likely to decline as mortality from IHD [ischaemic heart disease] continues to decline’. No one seems to have raised the possibility that heart disease rates have declined in the last 50 years partly as a result of rising alcohol consumption, nor was it pointed out that heart disease — declining though it may be — still kills more people than all the ‘alcohol-related’ cancers combined.

At this stage, however, there was little to suggest that the male drinking guidelines would be reduced to bring them in line with those of women. The Liverpool report had shown clear differences in risk for men and women, and a presentation viewed by the committee on 10 September 2013 showed different J-curves for both sexes. The draft conclusions of November 2013 noted that alcohol’s ‘cardioprotective effect on total mortality is observed at a much lower level of consumption for women’ and almost every country in the world has higher guidelines for men.

This began to change after the committee, via Public Health England (PHE), commissioned some computer modelling to help formulate the final guidelines. In the minutes of a September 2014 meeting it was noted that ‘the PHE tender exercise had resulted in just one bid and that the bidder would be interviewed in early October to explore their proposals’. We must presume that the lone bidder was Petra Meier and John Holmes’s team at Sheffield University since it was they who won the contract.

It remains puzzling why a theoretical model was deemed necessary when so much epidemiological data exists to show the effect on morbidity and mortality from different levels of alcohol consumption. We may never know how the model was put together — the published Sheffield report does not provide enough data to allow independent replication — but one thing is clear: its risk curves bear no relationship to any risk curves in the published epidemiological literature. Whereas observational epidemiology shows lower rates of mortality for people drinking up to 4-8 units per day, the Sheffield model suggests that drinkers’ mortality risk is lower than abstainers only at very low intakes and exceeds that of abstainers at around two units per day. Moreover, while epidemiological studies find that men can drink more than women before assuming the same risk as a teetotaller, the Sheffield report finds similar limits for both sexes.

It is not even clear what the Sheffield report is measuring. The key criterion for gauging a safe drinking level is the risk of death, ie mortality risk, but the Sheffield report instead focuses on mortality from ‘chronic alcohol-related causes’. It is obvious that non-drinkers are less likely to die from alcohol-related causes, but it tells us nothing about overall mortality.

Moreover, the team stripped out all health benefits from drinking with the exception of heart disease. In their response to a comment from the peer-reviewer (who, interestingly, said ‘I predict that there will be very little, if any, change to the guidelines’), the team stressed that ‘excepting cardioprotective effects, the report focuses exclusively on the negative consequences of drinking’. This was certainly true and, like Petticrew and Bellis, the Sheffield team went out of their way to cast doubt on the cardiovascular benefits. In the space of two sentences, they described the protective effect on the heart as ‘disputed’, ‘overestimated’ and suggested that the scientific consensus was moving towards the view that the benefits barely existed at all. This passage contains ten references, half of which were articles or op-eds written by Tim Stockwell.

In short, the Sheffield team produced a theoretical model that was entirely divorced from the epidemiological evidence. The model appeared to show that a ‘safe’ level of drinking — if defined as carrying no more risk than abstaining from drink — was significantly lower than had been reported in a large body of epidemiological research. The model also deviated from observational epidemiology by showing this ‘safe’ level to be similar for men and women alike. Indeed, it actually reported a higher level for women. When a computer model clashes with observed reality so conspicuously, it is time to bin the model. Instead, the CMO’s committee binned the real world evidence and used the model as the basis of its recommendations.

By April 2015, the only question was how to sell the new advice to the public. Having failed to completely erase the health benefits of drinking, the group were concerned about the public being encouraged to drink even small quantities of alcohol. The group emphasised that there was ‘now no justifiable case to recommend that anyone should choose to start drinking alcohol in the interests of their health’. Their message to those who already drank below the current lower risk limit was that ‘should they wish to reduce their frequency or levels of drinking, [they] need have no health concerns in doing so’. Whatever the evidence might say, there was no doubt that the committee favoured total abstinence: ‘The message is quite clear that any level of drinking can be harmful to health’.

The statistician David Speigelhalter, who acted as an adviser to the committee in the latter stages, told them that ‘a message that “there is no safe lower limit” would risk being at odds with public opinion’, but it barely seemed to matter if the public found the new guidelines credible or realistic. A telling comment in one set of minutes indicates that the real intention was to influence policy: ‘It would be important to bear in mind that, while guidelines might have limited influence on behaviour, they could be influential as a basis for government policies, which could in turn help to alter norms.’

The new guidelines were announced on 8 January 2016. The committee had made every effort to downplay the benefits of moderate drinking, and Sally Davies delivered the final blow by dismissing those benefits as ‘an old wives’ tale’ on the Today programme. Nearly three years after cramming her committee with temperance campaigners and ‘public health’ activists, Davies went further than even Tim Stockwell could ever have hoped when he had that first chat with Mark Petticrew in June 2013. The job was done.

Smoking and COVID-19: a new evidence update

Updated 16 December 2022 after a Facebook 'fact check'.

Studies which find smokers to be significantly less likely to get COVID-19 continue to be published, even if they are ignored by the media. There are so many of them that it's difficult to keep up. 

The indefatigable Phil has found over 4,000 studies in which smoking status is recorded (see this heroic thread), although the vast majority do not look at smoking specifically as a risk factor, so we can only compare smoking rates among Covid patients/cases with the smoking rate of the general population. Most of them seem to show smokers significantly under-represented among Covid patients/cases, but it is a crude method and is not considered conclusive (interestingly, however, the over-representation of fat people in Covid wards is considered sufficient evidence to mark obesity as a risk factor).

It is better to focus on studies which use epidemiological methods and adjust for other factors. There are quite a few of them now, including a growing number looking at antibodies in unvaccinated populations to see who has had Covid. These studies help address sampling bias by testing whole populations; they do not rely on people's willingness to come forward or on their ability to identify their symptoms. 

There is an ongoing meta-analysis, but it hasn't been updated since March (UPDATE: the seventh and final version has been published in Addiction). The main conclusion is that smokers are less likely to be infected and ex-smokers are more likely to be hospitalised if they catch the virus. The early finding that smokers are less likely to be hospitalised with Covid seems to be explained by them being less likely to be infected in the first place.

What follows is not a comprehensive list. These are just the studies that have crossed my radar. Please let me know about other evidence in the comments, including any studies that contradict the general conclusion that smokers are less likely to be infected with the virus. All of these studies have been published and peer-reviewed.

1. Lusignan et al. This British study published in the Lancet found that "active smoking was linked with decreased odds of a positive test result" with an odds ratio of 0.49 (0.34–0.71). 

2. Hopkinson et al. This British study found that smokers were 27% less likely to test positive for COVID-19 although you have to look carefully to find the evidence because the lead author is the chairman on Action on Smoking and Health and he buries it as much as he can. The odds ratio is 0.73 (0.65-0.81). 

3. Chen et al. This UK study didn't find an association between smoking or vaping and Covid infection but did find an association between vaping and smoking (i.e. dual use) and Covid infection (1.10 (0.89-1.36)). 

4. Williamson et al. Meanwhile this British study published in Nature looked at the likelihood of dying from COVID-19 and found that smokers were slightly more at risk, or slightly less at risk, or neither, depending on how the figures were adjusted.

5. Rentsch et al. This study from the USA looked at 3,789 US military veterans aged between 54 and 75 who were tested for COVID-19, of whom 585 tested positive. Smokers were 55% less likely to test positive, with an odds ratio of 0.45 (0.35-0.57). 

6. Ghinai et al. Studyy of homeless people in Chicago which found that smokers were less likely to be infected (0.71; 95% CI, 0.60-0.85). 

7. Lan et al. Study from the USA finds smokers 90% less likely to test positive. Odds ratio: 0.1 (0.01-0.8). 

8. Gu et al. American study found: 'Being a current smoker (self-reported in the latest EHR encounter) was associated with a reduced chance of having positive test results (OR, 0.31 [95% CI, 0.20-0.48]; P < .001).'

9. Vila-Córcoles et al. Study from Spain finds smokers 57% less likely to be infected (0.43 (0.25-0.74)). 

10. Miyara et al. French study finds that daily smokers are 76% less likely to be infected with COVID-19 (after adjusting for age and sex). Odds ratio for inpatients: 0.24 (0.14-0.40). For outpatients: 0.24 (0.12-0.48). 

11. Kharroubi et al. Study of workers in a public health laboratory in Tunisia found that "tobacco smokers had a lower risk of infection (AOR = 0.54 [0.29–0.97])".

12. Paleiron et al. Interesting study of a Covid outbreak on a French Navy aircraft carrier. 76% of the crew members got the disease but smokers were 36% less likely to get it (0.36; 95% CI, 0.49-0.81). There was also a "trend towards a lower risk among e-cigarettes users". Risk of infection was even lower for heavy smokers than for those who smoked less than 10 cigarettes a day.

13. Prinelli et al. Web-based study from Italy finds smokers were half as likely to get COVID-19 than nonsmokers. There was a dose-response relationship, with heavy smokers 62% less likely. 

14. Lee et al. Study of 4,137 Covid patients in South Korea finds smokers 67% less likely to be infected (0.33, CI = 0.28–0.38). 

15. Prats-Urbine et al. This study used UK Biobank data and found that current/former smokers are 45% more likely to test positive for SARS-CoV-2 (1.45 (95% CI 1.19 to 1.79)).

16. Fernandez-Fuertes et al. Study from Spain involving HIV patients. It found that 'active tobacco smoking was the only factor independently associated with lower risk of SARS-Cov-2 infection [Incidence rate ratio: 0.29 (95% CI 0.16–0.55)'.

17. Candel et al. Study of 10,614 nurses in Madrid finds smokers 77% less likely to have had COVID-19 (0.23 (0.20-0.27)). The finding is not mentioned in the abstract.

18. Cummins et al. UK study finds smokers are 27% less likely to be hospitalised with COVID-19 (0.63 (0.44-0.88)). The authors do not mention this finding in the text. 

19. Wratil et al. Study of German health workers finds smokers are around half as likely to have antibodies for SARS-CoV-2.

20. Hull et al. Study of ethnic minorities in London finds that smokers are 40% less likely to be infected (0.60 (0.56 to 0.63)).

21. Barchuk et al. Study from Russia finds smokers are 54% less likely to have had Covid (as measured by antibody tests).

22. Jose et al. American study finds smokers are 57% less likely to be diagnosed with Covid. Vapers were no more and no less likely to get the virus than nonsmokers.

23. Holuka et al. Study from Luxembourg finds smokers are 50% less likely to have had Covid.

24. Iruretagoyena et al. Study of healthcare workers in Chile finds smokers are 62% less likely to have had Covid. 

25. Li et al. Study in the Journal of the American Medical Association finds smokers to be 40% less likely to have had Covid.

26. Muro et al. Study from Spain finds smokers are 43% less likely to have had Covid.

27. Clift et al. Study from the UK finds no inverse relationship between having a genetic liability for smoking and SARS-CoV-2 infection. Naturally, this one got plenty of media attention. I wrote about it here. When using conventional epidemiology, the study found that "heavy smoking (≥20 cigarettes/day) was associated with a reduced risk of confirmed infection when adjusting for age and sex (OR 0.50, 95% CI 0.29 to 0.89)".

28. Maraqa et al. Study from Palestine finds that smokers are half as likely to have been infected with SARS-CoV-2 (0.47 (0.31-0.72)).

29. Duszynski et al. Study of 8,124 people in Indiana, USA, finds that smokers - but not vapers - are half as likely to get COVID-19 (0.49: 0.32-0.74).  

30. Pagani et al. Italian study finds that smokers are 58% less likely to get SARS-CoV-2 (OR 0.42: 0.29–0.60). Unusually, it also finds that former smokers are at reduced risk (OR 0.49: 0.33–0.75).

31. Mutevedski et al. Study from South Africa finds that daily smokers are half as likely to be infected with SARS-CoV-2 as non-smokers (0.50: 0.38-0.67).

32. Rosoff et al. Mendelian Randomisation study finds that smoking is not associated with lower risk of infection. Although the study is titled 'Smoking is significantly associated with increased risk of COVID-19 and other respiratory infections', the text makes it clear that the association is only with people with a "genetic liability for smoking". Strangely, it is only MR studies that find such an association with 'smoking' (see 27 above).

33. Egede et al. Study of 31,549 people tested in Wisconsin, USA last year finds smokers 66% less likely to have COVID-19 (0.34; significant at p < 0.001). Former smokers were 21% less likely. Those who did were more likely to be hospitalised. Smokers with Covid were no more likely to die of it than nonsmokers, but former smokers were.

34. Salerno et al. Study of nearly half a million dialysis patients in the USA finds that tobacco use is associated with a 16% reduction in risk of COVID-19 diagnosis (OR 0.84: 0.81-0.87).

35. Vallarta-Robledo et al. Study from Switzerland finds smokers are 56% less likely to be infected with SARS-CoV-2 after adjusting for potentially important confounders (OR 0.44 (0.35-0.77).

36. Modenese et al. Study from Italy finds smokers 30% less likely to be diagnosed with SARS-CoV-2 (0.70 (0.54–0.91)).

37. Ozcifci et al. Study of Behçet's syndrome patients finds smokers are 34% less likely to be diagnosed with SARS-CoV-2 (0.66 (0.47-0.93)).

38. Gashi et al. Study of municipal workers in Kosovo finds smokers are half as likely to have had SARS-CoV-2 (0.52, 0.28-0.97). 

39. Siller et al. Study from Austria finds that smokers are 61% less likely to have had SARS-CoV-2 than nonsmokers (0.39 (0.27-0.56)).

40. Madhi et al. Study in the New England Journal of Medicine looking at Gauteng, South Africa finds that daily smoking is associated with a 14% reduction in infection risk (0.86 (0.82-0.90)). Note that this is the first study of this kind involving Omicron. 

41. Verburgh et al.  Study from the Netherlands finds that smoking is associated with a 58% reduction in infection risk (0.42 (0.18-0.99).

42. Coppeta et al. Study of healthcare workers in Italy finds that smokers are more likely to get SARS-CoV-2 than nonsmokers (1.6 (1.1–2.4).

43. Zuñiga et al. Study from Chile find that tobacco consumption is associated with lower risk of SARS-CoV-2 infection, with effects ranging from 0.62 to 0.85. 

44. Constantino et al. French study of patients with rheumatic diseases finds that smokers are much less likely to test positive for COVID-19 (0.18 (0.03–0.61)).

45. Coric et al. Study from Serbia finds that smokers are 71% less likely to have COVID-19 (0.29 (0.15–0.55)).

46. Morino et al. Seroprevalence study from Spain finds that "users who smoked had a lower prevalence of [SARS-CoV-2] antibodies than non-smoking users, with a PR = 0.60 (0.39, 0.92)".

47. Warszawski et al. Study from France finds a "strong inverse correlation between the presence of SARS-CoV-2 antibodies and smoking... as in other studies". Nonsmokers were about twice as likely to test positive. 

48. Sandal et al. Study of Turkish healthcare workers finds smokers 62 per cent less likely to get COVID-19 (0.38 (0.23-0.63)). 

49. Rojanaworarit et al. Study from New York finds that current smokers are 40 per cent less likely to get COVID-19 (0.60 (0.49-0.74)). The authors note that "consistently low prevalences of smoking have been observed among patients hospitalized with COVID-19 in several studies. In an outpatient setting, smokers have been shown to have lower odds of SARS-CoV-2 test positive results".

50. Fuente et al. Study from Spain concludes: "Smoking was the only factor associated with a decreased risk of SARS-CoV2 infection of any grade [odds ratio (OR) 0.491; 95% CI 0.275–0.878; p = 0.017]."

51. Halili et al. Study from Kosovo finds lower rate of infection among smokers (0.56 (0.31, 0.99)), but the association is no longer statistically significant after adjustments (0.55 (0.31-1.01).

52. Vial et al. Seroprevalence study from Chile finds that "Smoking showed a negative association with SARS-CoV-2 infection." The authors note that "the protective effect of tobacco use has also been evidenced in European countries, reporting a significant negative association between smoking prevalence and the prevalence of COVID-19 across the 38 European nations after controlling for confounding factors (p = 0.001). Some authors propose that nicotine prevents infection by competing with the virus with the ACE2 receptor. In this sense, epidemiological and experimental evidence has been found." Indeed.

53. Richard et al. Study from Switzerland finds that smokers are 64% less likely to test positive for SARS-CoV-2 (0.36, 95% CI 0.23–0.55).

54. Friedman-Klabanoff et al. Study from the USA finds: "Smoking was associated with a lower risk of SARS-CoV-2 infection with a RR of 0.25 (95% CI: 0.06, 0.99)." 

55. Cherif et al. Seroprevalence study from Tunisia finds that "current tobacco smokers had lower SARS-CoV-2 seroprevalence than non-smokers (OR = 0.5 (0.4–0.6)".

56. Haller et al. Study from Switzerland looking primarily at face masks finds that the only factor that has a statistically significant negative association with COVID-19 is active smoking.  

57. Kahlert et al. Another study from Switzerland finds smokers to be 60% less likely to get COVID-19 than nonsmokers (aOR 0.4, 95% CI 0.2–0.7). 

58. Gornyk et al. Seroprevalence study from Germany finds smokers were 60% less likely to have had Covid (0.4 (0.3-0.7)).

59. Shafrir et al. Study from Israel finds smokers are 46% more likely to test positive (and are no more likely to suffer from severe disease if they do). 

60. Tang et al. Study from Canada finds smokers are less likely to be infected but the difference is not statistically significant.  

61. Villarroel et al. Study from Bolivia finds that "tobacco smoking was negatively associated with seropositivity (25.4%, OR: 0.420; 95% CI 0.24–0.74)".

62. Yeung et al. Another Mendelian Randomisation study (see 27 and 32 above) which found that people who have a genetic predisposition to smoke (but who do not necessarily smoke) are 19% more likely to get COVID-19 (1.19 (1.11-1.27)). My commentary on this method is here.

63. Gao et al. Massive study of nearly 8 million people in the UK finds smokers to be less likely to be hospitalised with COVID-19 and much less likely to enter ICU. Former smokers were more likely.

64. Basso et al. Study from Italy looking at breakthrough infections in vaccinated healthcare workers finds infections are 20% less common among smokers.  

65. Badial et al. Study from Switzerland finds smokers 70% less likely to get COVID-19 (0.3 (0.2-0.4)). 

66. Schonfield et al. Study from Argentina finds that "those with a history of smoking showed the lowest seroprevalence, a finding also reported in an Argentine case study and in other epidemiological studies."  

67. Zhong et al. Study from China finds that "smokers were less likely to develop COVID-19 (OR = 0.224, 95% CI = 0.084–0.592 p = 0.003)." This is a risk reduction of 78%. 

68. Erber et al. Study of hospital workers in Germany finds smokers are 48% less likely to develop COVID-19 (OR=0.52 (0.26–0.94)) (see table 2).

69. Shanaube et al. Study from Zambia finds that daily smokers are 53% less likely to have had COVID-19 (0.25-0.90). 

70. West et al. Swiss study of health and social care workers finds that smokers were much less likely to have had COVID-19 (OR 0.26; 95% CI 0.097–0.696).  

71. Nguyen et al. US study finds that "Smoking history, adjusting for medical conditions, appears to be protective for COVID-19 outcomes for <65 year olds". These outcomes include not only infection, but hospitalisation, ICU admission and (for those under 50) death. Above the age of 65, these outcomes worsened for smokers, although there was no increased risk of infection.   

72. Djukic et al. Study from Serbia finds that "smoking was associated with a decreased risk of COVID-19 development (OR = 0.22, 95%CI: 0.14–0.35, p < 0.001)."

73. Young-Wolff et al. Massive US study involving 2,427,293 people in California finds that smoking was associated with lower risk of infection, hospitalisation, ICU admission and death from COVID-19. "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." In other words, smokers were a third less likely to catch it and half as likely to die from it.

74. Kleynhans et al. Study from South Africa finds that nonsmokers are nearly four times more likely to get COVID-19 (OR=3.9). 

75. Tomaselli et al. Seroprevalence study from Italy finds that smokers are 77% less likely to have had COVID-19 (0.23 (0.12-0.45)). Smoking status was verified by testing for cotinine. 

76. Peremiquel-Trillas et al. Seroprevalence study from Spain finds smokers are 62% less likely to have had COVID-19 (0.38 (0.18-0.79)). 

77. Hausfater et al. Study of healthcare workers in France finds that "current smoking was associated with reduced risk (0.36 [0.21; 0.63])". 

78. Nernani et al. Study of long term patients in psychiatric hospitals in New York finds that smokers are 25% less likely to get COVID-19 (0.75 (0.60-0.90)). The authors do not mention this finding in the text. 

79. Ramirez et al. Seroprevalence study from Switzerland finds that smokers are 19% less likely to have had COVID-19, but the difference is not statistically significant (0.81 (0.53-1.22)). This finding is tucked away in the supplementary material (p. 6). 

80. Wessendorff et al. Study from Germany finds that smokers are 68% less likely to get COVID-19: "we observed a reduced risk of infection (OR 0.32, 95% CI 0.12 to 0.81) even after adjustment for ‘time spent outside’".

81. El-Ghitany et al. Seroprevalence study from Egypt finds that ("surprisingly") smokers were 81% less likely to have had COVID-19.

82. Zar et al. Study from South Africa finds smokers 57% less likely to have COVID-19 (0.43 (0.28-0.66)).

83. Huynen et al. Study from Belgium finds no statistically significant difference between seropositivity of smokers and nonsmokers (adjusted OR=0·43, 95% CI: 0·27; 0·69, p<0.001).

84. Wei et al. Study of essential workers in the USA finds that smokers were 73% less likely to be infected (0.27 (0.12-0.61)). 

85. Abdi et al. Study of HIV patients in South Africa finds that smokers are 43% less likely to have had COVID-19 (0.57 (0.36, 0.90)).

86. Leclercq et al. Study from Belgium finds that "being a smoker (OR 0.36 (95% CI 0.18–0.72)) was negatively associated with having... antibodies". 

87. Javed et al. Study from Pakistan finds that smokers were 45% less likely to have been infected with SARS-CoV-2 (0.550 (0.424–0.712)). 

88. Weber et al. Study of healthcare workers in France finds that smokers are 40% less likely to have had COVID-19 (0.6 (0.4-0.9)). 

89. Rahman et al. Study from Bangladesh finds that "smoking (OR 0.70; 95% CI 0.55 to 0.89) was associated with lower risk of seropositivity".

90. Laval et al. Study of an outbreak of COVID-19 on (another) French aircraft carrier finds that "smoking was associated with reduced infection (adjusted odds ratio (ORa): 0.57; 95% confidence interval (CI): 0.44–0.73)".

91. Colaneri et al. Study from Italy finds that "being a current smoker was negatively associated" with SARS-CoV-2 infection (0.43 (0.23–0.80)). 

92. Dadgari et al. Study from Iran finds no association either way between smoking and COVID-19 (1.08 (0.84-1.35)).

93. Mostafa et al. Study from Egypt finds positive associations with both former and current smoking which disappear after adjustment for confounding factors. Current smokers: 0.65 (0.38-1.09).

94. Dörr et al. Study from Switzerland finds that smokers are 32% less likely to be infected, but neglects to mention it in the text (OR = 0.68 (0.52-88)).

95. Alasmari et al. Seroprevalence study from Saudi Arabia finds that smokers are half as likely to have been infected (OR = 0.48 (0.29–0.78)).   

95. Ferrara et al. Study from Italy finds no statistically significant reduction in infection risk among smokers (OR = 0.61 (0.35-1.06)).

96. Örtqvist et al. Study of pregnant women in Sweden and Norway finds that: "Smoking was associated with decreased odds of test-positivity in women under non-universal testing (aOR 0.46, 95% CI 0.31–0.70), although this association was not observed in analyses of women under universal testing (aOR 1.06, 95% CI 0.72–1.56)". 

97. Dev et al. Study from India finds an association between smoking and infection but this disappears after adjustment. 

98. Adorni et al. Italian web-based study finds that 'current smoking (aOR 0.66, 95% CI 0.50-0.87) was associated with decreased odds' of having COVID-19.

99. El Moussaoui et al. Study of healthcare workers in Belgium finds that fewer smokers tested positive for SARS-CoV-2 but the difference was not statistically significant (0.57 (0.29–1.1)).

100. Zar et al. Study of mothers in South Africa finds that "current smoking was associated with seronegativity (adjusted OR=0·43, 95% CI: 0·27; 0·69, p<0.001)", i.e. the smokers were 57% less likely to have had Covid.

101. Whittemore et al. Study from New York finds that smokers/vapers were 17% less likely to test positive for COVID-19 (0.83 (0.80-0.87)). The finding appears in Table 2 but is not mentioned in the text.

102. Green et al. Study from Israel involving patients with bronchial asthma concludes: 'A significantly higher proportion of smokers was observed in the COVID-19–negative group than in the COVID-19–positive group (4734 [13.45%] vs 103 [4.55%]; P < .001).'

103. Gu et al. Study from Michigan finds that smokers were much less likely to test positive for COVID-19 (0.31 (0.20-0.48)).

104. Lombardi et al. Seroprevalence survey of Italian healthcare workers finds smokers were 59% less likely to test positive (0.41 (0.27-0.61)).

105. Holuka et al. Study from Luxembourg finds that smokers are half as likely to test positive for COVID-19 (0.50 (0.30–0.83)).

There is also an ecological study that compared smoking rates and COVID-19 mortality rates in 38 European countries and found an inverse relationship. 

There is also a large study from the USA which doesn't look at prevalence but, very interestingly, finds that smokers with COVID-19 who are given nicotine in hospital have a better survival rate and COVID-19 vaccines have more effect on smokers than on nonsmokers.

As mentioned above, there are numerous studies comparing the number of smokers in hospital with COVID-19 with the number of smokers in the general population. They mostly produce a similar finding to the majority of studies listed above. Some of these studies are well done - like this one from France - but I don't consider the methodology to be quite rigorous enough to justify inclusion in the full list. (Note that a simple comparison between the % of patients and the % of the general population was enough for people in public health to acknowledge that obese people were over-represented while the under-representation of smokers continues to be ignored.)

Finally, the UK's weekly ONS infection survey consistently shows that smokers about half as likely to catch the virus as nonsmokers. This is almost never remarked upon.

As for why this finding keeps emerging, opinions differ. One suggestion is that smokers spend more time outside, but this sounds rather like a cope given the size of the effect. Smokers do not spend all that much more time outside and it seems unlikely that this would give frontline healthcare workers, for example, significantly more protection.

It may be due to the nicotine, although it is notable that study 23 and study 29 did not find any benefits from vaping. This study discusses some possible biological mechanisms.

Various biological mechanisms have been proposed which are beyond my level of scientific understanding. In any case, the consistency of these findings is very interesting and the lack of interest shown towards them by the public health establishment is revealing.

Meanwhile, the legacy 'public health' research community is annoyed that people are freely sharing such information on Twitter. 

PS. This is too complicated for me to understand but it offers a scientific explanation for why smokers are less likely to get SARS-CoV-2.