The surprisingly controversial benefits of moderate drinking

The neo-temperance lobby can't handle the truth

(The Spectator has a rather annoying habit of deleting old articles from the internet so here is something I wrote for the October 2020 edition of Spectator USA (which I think is now known as Spectator World) and is no longer available online. It’s about drinking guidelines and the benefits of moderate drinking. I have added links and graphs.)

It was always likely that once the killjoys had done their work on smoking they would turn their attention to alcohol. Sure enough, with the Dietary Guideline Advisory Committee going through its twice-a-decade revision of what, and how much, Americans ought to be eating and drinking in order to look after their health, drinking alcohol is being subjected to the same demonization process that was once applied to smoking tobacco. There is a campaign to lower safe drinking limits in the US, in the same way that they have been lowered in other countries. Worse, there is pressure to eliminate altogether the concept of a ‘safe level’ of alcohol consumption — and make out that every drop brings a drinker a little closer to his or her demise.

There is just one little problem in the new temperance campaign. It simply isn’t true that a moderate level of alcohol consumption is harmful. On the contrary, there is a lot of evidence to suggest that a drink or two a day will reduce your risk of an early death. The health benefits were first observed by scientists in the 1920s and identified in epidemiological studies in the 1970s. The relationship between alcohol consumption and mortality is J-shaped: the lowest levels of mortality are found among people who drink moderately. Drink heavily and you will significantly increase your risk of dying. Abstain altogether, on the other hand, and you will also increase your risk of dying early, albeit not by as much as by drinking heavily.

The J-curve first appeared, to my knowledge, in a 1994 study by Sir Richard Doll that employed the same method of surveying doctors that he had used 40 years earlier to show the association between smoking and lung cancer. Replicated many times since, Doll’s graphs (shown below) suggested that individuals had to drink a considerable amount of alcohol before their mortality risk — and their risk of death from heart disease, in particular — exceeded that of someone who doesn’t drink at all.

Correlation, of course, does not equal causation and Doll’s study could not prove that alcohol consumption explained the differences in mortality. Alternative explanations were put forward, including the possibility that some nondrinkers may have been former drinkers who had damaged their health and were therefore at greater risk of premature mortality. This came to be known as the ‘sick quitter’ hypothesis.

As he had done when researching smoking, Doll tested such hypotheses. A few months before his death in 2005, he published another study based on 23 years of data which replicated the results of the previous research and disproved the sick quitter hypothesis by comparing lifelong non drinkers with moderate drinkers. It found that moderate drinkers had a 19 percent lower mortality risk and a 28 percent lower risk of ischemic heart disease. By this time, Doll had concluded:

‘That the inverse relationship between ischemic heart disease and the consumption of small or moderate amounts of alcohol is, for the most part causal, should, I believe, now be regarded as proved.’

This evidence has continued to grow ever since. In recent months, new studies in Japan, Finland and the Netherlands have confirmed the J-curve. And yet there is a reluctance among some public-health academics and campaigners to accept the findings. In recent years, the J-curve has come under renewed attack, driven more by politics than by any change in the evidence base. The benefits of moderate drinking are an inconvenience to those who want to introduce restrictive tobacco-style legislation for alcohol. If alcohol can be part of a healthy lifestyle, it is difficult to demonize. Thus, for the new breed of temperance campaigner, it would be better if there were no safe level of alcohol consumption.

The way in which national drinking guidelines are established should be an arcane scientific process of little interest to the general public, yet instead it has become a heated battleground. The year after Britain’s safe drinking guidelines were devised downward in 2016, I carried out an investigation into the process. Through an analysis of the guidelines committee’s documents, minutes and emails released under the Freedom of Information Act, I discovered that the process had been captured by anti-alcohol activists and that modeling commissioned from the University of Sheffield was changed at the 11th hour at the insistence of the funder, Public Health England (PHE).

The original modeling implied that no change to the guidelines was needed, but the revised modeling suggested that the safe drinking level for men should be reduced. The switch in methodology demanded by PHE had no scientific justification, as the Sheffield team admitted in a previously undisclosed email, but it was made nevertheless. When the new guidelines were announced in 2016, Britain’s chief medical officer went an outrageous step further by declaring the benefits of moderate drinking to be ‘an old wives’ tale’. Earlier this year, I submitted this evidence to the Australian government, which had employed the same Sheffield team to help it revise its own drinking guidelines. Yet in Australia, the guidelines ended up being revised downward.

For the anti-alcohol lobby, lowering drinking guidelines serves a useful propaganda purpose — it creates a large number of new hazardous drinkers overnight, thereby inflating the scale of the problem. In countries such as Britain, where the proportion of the population consuming alcohol above the recommended amount has been falling for years, lowering the threshold boosts the figures and disguises the progress that has been made. This is acknowledged by the Sheffield modelers themselves, who say that ‘Advocates leverage the consumption thresholds in public debate to illustrate levels of excess alcohol consumption and the need for intervention.’

The alarming news that more people are apparently drinking at risky levels acts as a spur to lawmakers to take tougher action on alcohol. The latest report from the Dietary Guidelines Advisory Committee mentions ‘prioritizing policies that may lead to changes in consumption’ as a reason for lowering the guidelines. In its draft scientific report the committee states that ‘risk starts to increase above the equivalent of one-half US standard drink per day on average for women, above one-half to one drink per day on average for men’. Yet it fails to present specific citations to support this conclusion and, indeed, admits that very few studies have examined these ‘finer gradations’ of alcohol consumption. In fact, the report concedes that just one study ‘examined differences in risk amongst men consuming 1 v 2 drinks per day’.

The report does at least acknowledge that not drinking at all is associated with higher risk than drinking moderately. But although a guideline of one drink per day for men might reflect the optimal level of drinking, it would not reflect the ‘safe’ level — a reasonable definition for which is the level at which mortality risk begins to exceed the risk for nondrinkers. The committee argues quite vehemently that alcohol is inherently unhealthy and those who do not drink alcohol should not begin to drink for any reason. But unless the US government is going to advise nondrinkers to become light drinkers — and it clearly is not — there is no justification for deriving guidelines based on the optimal, as opposed to the safe, level.

The committee’s analytical gymnastics go to great lengths to avoid the obvious conclusions confirmed by a wealth of epidemiological evidence. These studies have been reviewed and aggregated in meta-analyses many times over. One such meta-analysis, by Castelnuovo et al. (2006), defined a ‘drink’ as 10 grams of alcohol (a standard drink in the US contains 14 grams) and found a protective effect from drinking two to four drinks a day for men and one to two drinks a day for women. Another meta-analysis by Jayasekara et al. (2014) found reduced risk among people drinking between 1 and 29 grams per day — i.e. an upper limit of two US drinks per day. A large study by Klatsky and Udaltsova (2007) used the US definition of a standard drink and found ‘increased risk for persons reporting more than three drinks per day and reduced risk for lighter drinkers’.

Recent studies by Xi et al. (2017) and Wood et al. (2018) again point in the same direction. The former took a very large sample of US drinkers and compared them to a large number of lifetime abstainers. It found a significant reduction in all-cause mortality among drinkers consuming up to 14 drinks a week (men) or up to seven drinks a week (women), with reductions of up to one-third in cardiovascular mortality. Wood et al. analyzed 83 studies and found that a drinker’s mortality risk only began to exceed that of a lifetime abstainer at 300 grams per week, equivalent to 3.4 US drinks per day. Their findings (shown below) once again confirm the J-curve.

Taken together, this considerable evidence suggests that it is certainly safe for men to consume up to two standard US drinks per day and arguably more. In fact, a strong case could be made for raising the guidelines for both genders. One factor which tends to be overlooked is that all epidemiological studies rely on people accurately reporting their own level of alcohol consumption. Yet it is well known that people greatly underestimate how much they drink. Surveys which compare the amount people say they drink with the volume of alcohol actually sold suggest that drinkers report only 40 to 60 percent of what they consume — either that or they are pouring an awful lot down the drain.

If we accept there is massive and consistent under-reporting of consumption, we can say with confidence that the risks associated with a given amount of drinking are lower than they appear to be. People who say they consume one drink per day are actually consuming around two drinks per day. People who say they consume two drinks per day are actually consuming around four drinks per day, and so on. Therefore, the risk ratio reported in epidemiological studies for someone who says they drink two standard drinks a day is actually the risk ratio for someone who consumes around twice that amount. This issue is rather obvious and yet it gets remarkably little attention, presumably because it implies that guidelines should be doubled, not halved.

With 30 years of epidemiology under its belt, the J-curve seems to be proven beyond reasonable doubt. Nevertheless, a handful of activist-academics continue to cast doubt on the evidence. Much of this revolves around ‘confounding factors’. The Dietary Guidelines Advisory Committee report, for example, cites a study which found, not for the first time, that people with low socioeconomic backgrounds — another factor associated with early death — are more likely to be lifetime abstainers. Another study found that lifetime abstainers are less likely to be highly educated, and a third found that young adults with a long-standing illness were less likely to be drinkers.

This is all true, and these variables could affect a study’s results. Yet epidemiological studies do try to adjust for these factors in the same way that they adjust for age, socioeconomic characteristics, race and so on. Moreover, there are also confounding factors which are likely to exaggerate the ill effects of high alcohol consumption. Heavy drinkers are much more likely to smoke, take risks and be impulsive, for example.

There is a double standard at work. Public health agencies are quite content to accept the results of observational epidemiology when studies show a link between alcohol consumption and, for example, breast cancer. No one talks about confounding and ‘healthy abstainers’ when alcohol is linked to any number of diseases, often with far weaker evidence than supports the indomitable J-curve. But when it comes to the benefits of moderate drinking, only a randomized controlled trial would satisfy the skeptics — and they know that such an experiment would be very difficult to conduct.

The latest glimmer of hope for those who wish to erase the J-curve is Mendelian Randomization (MR), a technique which seeks to screen out confounding factors in epidemiological studies by grouping people who have a genetic disposition towards particular types of behavior — for example, there is a gene, ALDH2, which is associated with heavy drinking.

There are MR studies that have produced an interesting new development in this field of research, and several involving alcohol have received a huge amount of attention, but they have at least as many problems as observational epidemiology. In a commentary in the European Journal of Epidemiology, Kenneth Mukamal and colleagues argued that MR ‘is subject to all of the limitations of instrumental variable analysis and to several limitations specific to its genetic underpinnings’. MR studies on alcohol consumption and cardiovascular disease, they write, ‘demonstrate that it must be treated with all of the circumspection that should accompany all forms of observational epidemiology’.

MR often does not measure alcohol consumption at all. It merely assumes that people with certain genes will drink less than other people. But just as someone whose genes are supposed to make him or her more likely to smoke will not necessarily become a smoker, someone who has genes that make them more prone to alcohol avoidance is not necessarily a non-drinker (or even a light drinker).

The neo-temperance lobby was delighted when a few MR studies failed to find a protective effect from moderate drinking on heart disease, but said nothing when the same kind of studies found no association between alcohol use and cancer. The epidemiological evidence linking alcohol to several forms of cancer is nearly as strong as the evidence linking moderate drinking to lower rates of heart disease, and yet MR studies have not yet been able to replicate it. If we allow MR studies to override epidemiological findings, several babies will have to be thrown out with the bathwater.

Taken in excess, alcohol is linked to a number of negative health and social outcomes. That is sufficient to justify extra regulation, including excise taxes and a ban on sale to minors. But there is no justification for treating alcohol like cigarettes, with exorbitant taxes, heavy regulation and the stigmatizing of consumers. Health advice should be credible and honest. It should reflect the evidence and should not be mangled for political reasons or bent to manipulate public opinion. We have a right to know the truth, and that is that moderate drinking not only does us no harm — it is good for us.