There is no safe level of drinking, according to a major new study. But who’s to say the very small risks of moderate drinking aren’t worth it?by Michael Blastland / December 10, 2018 / Leave a comment
It is the season of the handbrake turn: December for drink and indulgence; January for dry resolution. Ever-more health conscious, ever-tempted to treat ourselves, we spin from “what-the-heck” to self-discipline, caught between thoughts of pleasure and harm, never clear where the balance lies.
These feelings are shared even by Sally Davies, the nation’s Chief Medical Officer. She freely admits to enjoying a glass of wine, and yet says that every time she has an alcoholic drink she thinks “cancer.”
None of which is helped by the fact that the story of what’s known about the effects of drinking is disputed. Not at the higher end, of course. In quantity, there’s never been much doubt: it’s bad for you.
It’s at the lower end that arguments rage. And here, the story is still unfolding, but perhaps towards a kind of resolution. A word of warning: before we reach that resolution, things will get murkier. All this means that the next chapter could be the most intriguing of all, partly for what it tells us about the merciless difficulty of finding out the truth about everyday behaviours of all kinds, but above all for what it tells us about risk and the way we reckon with it.
On the gloomy side of the argument, there’s the expert advice to the recently-revised UK guidelines. This included the line: “There is no level of regular drinking that can be considered as completely safe,” a line we’d grown used to about smoking, but not about booze. The phrase “no safe level” made it into the guidelines only in relation to cancer and (controversially) for pregnant women. But that was enough, and, in the media at least, the phrase stuck.
Then, a few weeks ago, it was echoed by a study called “The Global Burden of Diseases,” published in the Lancet: “The health risks associated with -alcohol are massive,” the lead researcher said. “Zero alcohol consumption minimises the overall risk of health loss.”
All of which sounds conclusive enough. And yet, on the other side, isn’t there supposed to be something about moderate drinking being good for the heart and actually prolonging life?
There is. Quite a lot in fact. Study after study has found that if we observe how much people drink, then see how long they live, moderate drinking is associated with longer life than drinking nothing. In fact, not drinking came out in some studies as bad as a weekly skinful of 40-plus units, or about 20 pints.
Confusing, isn’t it? Is there no safe level? Or is there a positively benign level?
What doesn’t kill you
Naturally, these studies adjusted for other influences on health—so-called “confounders,” such as social status, which might distort the evidence. This will be a problem if, for example, the privileged and educated have more contented lives and better diets while also tippling a bit more than the poor.
Still, even with adjustments for such things, low levels of drinking came out beneficial. The results form a famous “J-curve”: higher risk in those who drink nothing, lower for those who drink a little, rising highest as the units really begin to stack up. The usual explanation for these findings is that while alcohol at any level is bad for cancer (hence, “no safe level”), small quantities are good for the heart, and in these moderate quantities the benefit to the heart wins out.
One remaining objection to those results is known as the “sick-quitter hypothesis”: some abstainers are former alcoholics or stop because of ill-health—and this is why they die earlier on average. But that objection seemed answered by studies that looked only at “never-drinkers” among the non-drinkers, thereby excluding those who gave up for health reasons. Still the J-curve applied. Even never-drinkers seemed on average to die earlier.
At this point, you might wonder why anyone—other than moralising temperance campaigners—doubted that a small amount of alcohol, on balance, does good. But it is still not ridiculous to worry that an overlooked confounder could lurk in these results. An underlying concern with all observational studies is that we can only control for confounders we can think of—like income or “sick quitters”—and only then if we observe and measure them properly.
Notoriously, big observational studies of the effect of vitamin E on heart disease once suggested that it reduced the risk by 40 per cent, a massive benefit against a common killer. That is, until it was investigated by experiment instead of observation. A randomised controlled trail (RCT) gave vitamin E to one group but not to another and the difference in heart disease disappeared. There was no value to the heart from these supplements. That a huge benefit could vanish once tested in an RCT is a measure of how hard it is to be sure that observational studies are not in some way residually skewed. Clearly, the kind of people who ordinarily took vitamin E were different in some other way that significantly helped—but, whatever it was, it was missed.
Did a similar, residual confounder lurk in those many observational studies of alcohol? Perhaps that those who drink moderately are moderate in all things, and healthier for it? The problem was how to find out. We couldn’t easily experiment on people in a long-term RCT and tell them how much to drink. We were stuck.
The sober alcoholic
But perhaps no longer. Evidence has recently emerged from a new source on the question of how drinking affects the heart. This research is not widely known, which is odd. Especially as the story of how we found out is every bit as intriguing as the answer itself.
Prior to his appearance, it was thought no such person could exist. Sadly, we don’t even know his name. The literature reveals little about him, except that he was Han Chinese and had the full set of genetic variants that make it physically horrible to drink. In particular, he had homozygosity (an allele from both parents) of the variant ALDH2*2.
These variants compromise the liver’s ability to break down alcohol, allowing more of the toxin acetaldehyde into the bloodstream. One of the more conspicuous reactions is sudden facial flushing. Others include nausea, headache and tachycardia. But what made him truly unusual was that despite this vicious biological discouragement, he still, somehow, became an alcoholic.
It took some doing. His strategy was to sip beer all day long for a daily intake of three to five small bottles at 4.5 per cent. No wine or spirits, no bingeing, just periodic sips, all day; the only way, presumably, to get even this much down. Yet try as he might, this man—a diagnosed alcoholic—could not get drunk. He was bashing his head against the wall of this genetic restriction—and the genetic restriction held up.
He is living evidence that ALDH2*2 truly is the most-powerful genetic modifier of alcohol consumption yet discovered—such that even an alcoholic cannot become intoxicated. He thus shows us why the millions of East Asians who have these variants drink far less than others and often abstain entirely. It also gives us confidence that different severities of genetic variant will be strongly associated with different levels of drinking.
What makes this valuable to the rest of us is that we can use it to investigate that most vexed question about moderate drinking. The genius of the idea is the realisation that nature has all the while been conducting something akin to a randomised trial of alcohol consumption.
It works like this: Gregor Mendel’s second law states that each gene is inherited independently of other genes. This means people acquire these variants at random. Because the effect of the variants is demonstrably powerful in this case, we can therefore identify a randomised low and non-drinking population from their genes. Thus what happens on average to the health of people with and without these genes can act as a proxy for the health effects of alcohol, untainted by confounding effects in people’s lives or environment.
The principle of using gene variants as proxies for environmental influences is known as Mendelian Randomisation (MR). It has great potential as a research tool and is becoming a very big thing. So what does MR tell us about the health effects of low levels of drinking on the heart?
It’s bad for you. The more restrictive the variant, the better the health outcomes. Any quantity of alcohol is probably detrimental on average.
Some people have reservations about MR: it can be done badly and requires certain assumptions, particularly that the genes in question do not confer any other health-enhacing effects. Still, its advocates are convinced. Michael Holmes, a researcher at Bristol’s Medical Research Council Integrative Epidemiology Unit has said: “We have decades of observational studies which consistently show that moderate consumption of alcohol is associated with lower risk of heart disease. The question has always been ‘is that a causal effect or is it due to confounding?’ We can show that in contrast to being protective, in actual fact consumption is linked to higher risks of heart disease.”
No safe level?
Let’s say, for the sake of argument, that it’s settled. Some will dash to claim that this tells us all we need to know: alcohol equals harm. One reaction to the latest evidence would be to reiterate the advice that there is, after all, “no safe level” and we should therefore consider abstention.
But that hasty conclusion overlooks two simple yet important points. First, that safety is merely the beginning of the calculation. As the statistician David Spiegelhalter has remarked: “there is no safe level of driving,” but no one argues on those grounds for abstention. Come to think of it, he says, “there’s no safe level of living,” and we don’t abstain from that either. Which is to say that we need a proportionate sense of potential harm and benefit/pleasure.
The second point follows: yes, harm is indubitably part of the calculation; but how likely is it for you? Harm on average clearly does not mean universal harm. Some will be affected, some won’t. What are your chances?
As usual in seeking proportion, there’s a hitch: Mendelian Randomisation suggests a direction for the effect (it’s bad for you), but not the size of the overall effect (how bad, exactly). No one has (yet) incorporated this new evidence into estimates of all-cause mortality or the full range of morbidities. Even that downbeat recent study in the Lancet assumes a benefit to the heart—though a small one outweighed by other malign influences. Then how would we get a handle on the risk of moderate drinking if we accept the new evidence? Not easily. Either we wait, or we’re going to have to guess. Seriously, that is currently the best we can do.
And that brings the next surprise: we can make a range of guesses that assume a more detrimental effect from low levels of alcohol than previously claimed, and still struggle to make it at all likely that you personally will be harmed.
Let’s begin our guesswork with the Lancet study. This found that among 100,000 people drinking one small glass a day (1.25 UK units), we’d expect 918 cases each year of the kind of illnesses caused by alcohol. In 100,000 people abstaining, we’d see 914 cases—a difference of four in 100,000.
Another way to think about these numbers is to imagine that you had 25,000 possible futures each year as a one-glass-a-day drinker. In one of those futures you’d suffer an alcohol-related disease that you could avoid by being teetotal. In the other 24,999 possible futures, you’d see no difference.
What happens if we raise that risk after taking a more pessimistic view of the effect on the heart. Well, if we just about double it, take that risk every year for 60 years, make every risk a risk not just of illness, but death, then still I reckon you’d be looking at a lifetime chance of dying early comparable to the risk of dying on the roads. And that’s with a lot of “ifs.” Is that a big risk, for you?
I emphasise the subjectivity because next, we come to the benefit, and another hitch: we have—and can have—no objective or standardised unit of pleasure. Philosophers once suggested such a unit and called it the Hedon. Unsurprisingly, it didn’t catch on. Does pleasure beat risk, with all the uncertainties? Only you can decide. Pleasure is a devil to nail down, but clearly part of a calculation which becomes both subjective and highly uncertain.
The fallacy of the average
My slightly heretical conclusion after going through the exercise of playing with the numbers is how little difference it makes to the uncertainties for any particular individual—a conclusion that applies whether we keep the J-curve or junk it in favour of “no safe level.”
The binary allure of “is it bad for you, yes or no?” calls for a belief that whatever we find to be true for a whole population on average will be true for you personally. That is a fallacy. There is a real but small chance of harm for anyone, but it is awash in a sea of uncertainty and other considerations for any plausible range of numbers around low-level drinking.
Those numbers hide further soft complications: factors like social pressure, guilt, and moral attitudes to drink, plus a tendency to lie about how much we consume which adds a whole new layer of uncertainty about the effects. This blend of many fogs will strike you differently on different occasions (December and January being two of them). All of which helps explain why a) it’s just about impossible for anyone to give you advice you’d be consistently happy with, including yourself, and b) trying to reduce risk to a mathematical calculus alone is pretty much doomed.
John Holmes of the University of Sheffield’s alcohol research group told me: “I agree that the uncertainties in the evidence should be highlighted more in public debate and these uncertainties extend well beyond those around heart disease and the J-curve. However, I am less sure how to communicate uncertainty to the public without adding to the confusion that already exists.”
I understand the concern that uncertainty can be taken too far. Some people abuse it as an excuse to believe whatever they want to. But uncertainty in this case doesn’t mean complete ignorance and carte blanche. Far from it. We can put rough numbers on the risk-side of the equation, and at the higher end of consumption, the chance of harm rises fast—to our peril. With that essential caveat, my view about uncertainty is that if the evidence is uncertain, people ought to be uncertain. Simplifying for convenience is what fairy tales do. Grown-ups need to grow up.
Admittedly, this makes issuing guidance tricky. I sympathise with those who struggle to frame it. “It’s bad for you!” would be easy to put on a tablet of stone, and might be true. “It’s uncertain for you!” doesn’t have the same ring. The problem is, that’s true too, and always will be. As is “and you might like it.” So good luck with all that. Because, in the end, risk is personal.