Moderate drinking and its enemies

A little bit of what you fancy does you good

More than any other academic, Prof Tim Stockwell has provided the studies that have given ammunition to the anti-alcohol lobby over the last 20 years. He was involved in lowering the drinking guidelines in the UK and elsewhere, he provided evidence from Canada that purported to show that minimum pricing works, and he continues to make an almost fanatical effort to cast doubt on the health benefits of moderate drinking.

Stockwell was interviewed by the knowledgeable Felicity Carter in a recent episode of the Drinks Insider podcast. He did his best not to sound like a zealot but I do not believe him when he says that he would love to show that moderate drinking is good for your health, nor do I believe him when he says that he is eager to find evidence that it is. I think he gives himself away with his virulent anti-industry rhetoric and I think that he finds the benefits of moderate drinking to be an obstacle to the neo-temperance legislation that he openly supports. His minimum pricing research was junk science on meth and I think he will use every statistical trick in the book to reach his desired conclusion. This became glaringly obvious earlier this year when he whittled 123 cohort studies down to five - based on opaque and seemingly indefensible selection criteria - and applied numerous, vaguely described adjustments to the data so that he could go to the media and say: “There is simply no completely ‘safe’ level of drinking.”

In the podcast, Stockwell acknowledges that most studies find a J-Curve like the one shown below, with teetotallers having a higher risk of premature death than moderate drinkers and heavy drinkers having an even higher risk. Studies of heart disease show an even deeper J-Curve, and it is the reduced risk of heart disease that is the main driver of lower mortality among moderate drinkers.

Stockwell has made it his mission since 2006 (when he was president of the neo-temperance Kettil Bruun Society) to find reasons why this effect is not due to alcohol consumption. He accepts that teetotallers are at greater risk, but says this is for one of two reasons.

1. The sick quitter hypothesis

If lots of people stop drinking due to ill health (regardless of whether their ill health was caused by alcohol or not), teetotallers could appear to be at higher risk of death than moderate drinkers, but this doesn’t mean that moderate alcohol consumption confers health benefits. This explanation for the J-Curve was first proposed in the 1980s and has become a favourite of mid-wits ever since.

This theory is fairly simple to test. All you have to do is exclude the former drinkers and see if people who have never drunk are more or less healthy than people who drink moderately. The renowned epidemiologist Richard Doll did this in one of his final studies, published in 2005, and found a 28% reduction in ischaemic heart disease among moderate drinkers and a 12% reduction in all-cause mortality.

Countless other studies have done the same thing in the last 20 years, with much the same conclusion. As the graphs below show, ex-drinkers do have a somewhat higher risk than lifetime abstainers, but both of them have a substantially higher risk than moderate drinkers.

2. The sickly abstainer hypothesis

The sick quitter theory has been so thoroughly debunked that I was surprised to find Stockwell still going on about it on the podcast. When pushed, he switched to what might be called the sickly abstainer theory. The claim here is that people who never drink alcohol are somehow predisposed to poor health or are born with health conditions that predispose them to not drink alcohol.

The idea that people who never drink are inherently unhealthy is dubious on the face of it. A very large number of people do not drink for religious reasons, most obviously Muslims and adherents to various Protestant sects. Many others do not drink because they do not like the taste. They are just normal people who have made a choice. There is no reason to think their lifestyles are less healthy than average and in some ways the lifestyles of the religiously inclined are likely to be healthier.

Stockwell says that someone born with a serious health condition is less likely to ever drink alcohol. No doubt that is true. He also says that people on low incomes are, on average, more likely to be lifetime abstainers than people on higher incomes, and that people on low incomes tend to die sooner. This is also true but, at root, this is a simple question of confounding and both of these confounding factors can be easily adjusted for. Epidemiologists control for socio-economic status all the time, just as they adjust for age, smoking, weight and many other factors. Epidemiological adjustments are not an exact science by any means, but it is easy enough to compare a bunch of low income never-drinkers to a bunch of low income moderate drinkers. As for never-drinkers who abstain from alcohol because they are born with debilitating health conditions, there are surely not enough of them to have any meaningful effect on the results of studies looking at hundreds of thousands of people.

The possibility that the J-Curve could be the result of confounding factors was raised decades ago, notably by A. G. Shaper (whose ideas were resurrected by Stockwell in 2006), and Richard Doll again responded by doing the appropriate statistical tests. By 2002 Doll had concluded that:

Confounding, often a serious concern when risks vary by less than two-fold, has in this case been tested and found wanting... 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.

There are no potential differences between never-drinkers and moderate drinkers that cannot be controlled for in an epidemiological study. In the podcast, Stockwell mentions a study by his longtime collaborator Tim Naimi (one of the brains behind the two-drinks-a-week health guidance in Canada) which found that teetotallers were more likely than moderate drinkers to have all sorts of risk factors for heart disease (one of the major exceptions was smoking). The study did not distinguish between never-drinkers and ex-drinkers, and all of the risk factors could be controlled for by a careful epidemiologist. Moreover, most of the risk factors were markers for poor health in general rather than just heart disease. But here’s the thing. If teetotallers were inherently less healthy we would expect them to do worse than moderate drinkers across the board and there should be a J-Curve for nearly all diseases (the exception, perhaps, would be liver disease).

But there isn’t. We see a J-Curve for cardiovascular disease, diabetes and dementia (and, more contentiously, colorectal cancer). We do not see a J-Curve for the vast majority of cancers. Doesn’t that seem strange if never-drinkers (and ex-drinkers, for that matter) are so unhealthy, deprived, fat, physically inactive and poorly nourished - or if they have some mysterious confounder that no one has yet identified which predisposes them to ill health?

Furthermore, we have plausible biological explanations for moderate drinking reducing the risk of the diseases for which there is a clear J-Curve. Surely even Stockwell does not deny that moderate alcohol consumption increases HDL (“good”) cholesterol concentration? If so, wouldn’t it be odd if moderate drinkers didn’t have lower rates of heart disease?

If someone were to propose the ‘sick smoker’ hypothesis, they could claim that smokers are (to use Stockwell’s phrase) “a funny lot”. They are more likely to take drugs, have unprotected sex, drive dangerously and eat badly. They are much more likely to live in low income households. All this is true. On that basis, it could be claimed that their higher rates of mortality are not due to smoking but to all their other bad habits and disadvantages.

This might sound compelling to an idiot, but it doesn’t take much to run a coach and horses through it. Firstly, smokers have not always been disproportionately poor. In the 1960s, blue collar workers were roughly as likely to smoke as white collar workers and yet smokers in general were dying at a higher rate than nonsmokers. Secondly, smokers are much more likely to get lung cancer and that has little if anything to do with diet, drugs, physical activity or any other behaviour. And it is the lung cancer risk than drives the mortality risk. In any case, epidemiologists can compare smokers who are poor and obese (or drug users or problem gamblers or whatever) to nonsmokers who are poor and obese (or whatever) and still see a clear difference in mortality rates.

Anyone who made started talking about sick smokers would be quickly recognised as a barrel-scraping crank. So it is with those who persist with the sick quitter and sickly abstainer theories. We have a leading cause of death (heart disease) that is very plausibly mitigated by consuming a certain quantity of alcohol and which can be expected to lead to a reduced risk of overall mortality for those who consume it. Drinkers, ex-drinkers and never-drinkers may be different in various ways, but normal epidemiological practice can deal with that. Epidemiology has had 40 years to deal with it and the J-Curve has stood up to everything that has come its way.

I am no sucker for observational epidemiology but I trust the people who conduct the studies to control for confounding variables more than I trust a psychologist with a bee in his bonnet.

You can listen to my own interview on the Drinks Insider podcast and there’s a good episode of The Studies Show podcast looking at the J-Curve here. (One of the hosts of the latter is a teetotaller and presumably destined for an early grave.)