Would you trade 1.8 years of life for a drink per week?
The co-author of a study that quantified drinking's impact on lifespan re-evaluates its conclusions.
Here’s the BLUF (bottom line up front):
∆ Regular alcohol use likely shortens lifespan; by how much is an educated guess (see below for as thorough a quantitative analysis as you will likely find).
∆ The consistency and duration of your drinking (in years and decades) significantly influences the impact on your lifespan.
∆ If you have 1-2 drinks per day for most of your adult life, research (see below) suggests the impact on your lifespan may be somewhere between 8 months and 2.8 years. But it could be less or more, depending on multiple personal and other variables.
∆ If you stop drinking, the length of that abstinence influences lifespan, e.g., quitting alcohol halfway through adulthood would likely reduce life lost lifespan from very roughly 2.8 years to very roughly 1.4 years.
∆ The research doesn’t account for any social, emotional or other enjoyment benefit (if that’s the right phrase) you may get from drinking. So just make informed choices.
Dear reader: If you like this post, please “like” this post! It really helps. Thank you.
When I published “Aging with alcohol,” which mentioned a study that included the spit-out-your-coffee conclusion that moderate drinkers lose 1 to 1.5 years of lifespan per drink per week, I also reached out to the study’s co-author to ask a question:
How could that ratio be accurate?
At 1 to 1.5 years per drink per week, a mere 5-per-week drinker would sacrifice 5 to 8 years? And someone who regularly consumes 10 drinks a week — not binge drinking, mind you — would lose an incredible 10 to 15 years of lifespan?
The numbers seemed unrealistic.
After nearly a dozen emails and phone calls attempting to reach Dr. Mary Schooling, the study’s co-author, she responded with two emails, each providing a gracious and detailed explanation and re-evaluation of her study’s conclusions.
Dr. Schooling, a professor at the City University of New York (CUNY) School of Public Health, provided context for her research findings and offered what a refinement of how much lifespan even moderate weekly drinkers sacrifice as a result of drinking. Her assessment: Moderate, decades-long moderate weekly drinkers probably lose between a year and two years of lifespan per drink per week. But then again, given all the variables, she seemed to suggest, even that estimate could be too aggressive.
Her detailed email replies to my two rounds of questions appears below. Warning: they get a little technical.
I’ve condensed her email for clarity and brevity and bolded certain phrases for emphasis. I have not otherwise altered her language in any way that would change the intent of her words.
Round 1: Here’s Dr. Mary Schooling’s first email reply, on Feb. 17, to my initial questions:
Hi,
Thank you very much indeed for your questions. You have asked a tricky question, which I have answered in detail. Please let me know if you have any further questions.
Our study is using genetic variants as proxies for alcohol use to reduce bias. The study is in people of European descent from Europe and North America. As genetics are lifelong, the estimates are in terms of lifelong alcohol use. The overall estimate we gave was 1.09 years of life lost per logged alcoholic drink per week. On reflection we should have said we meant regular drinking of this amount per week.
The logged alcohol drinks per week can also be converted into units of alcoholic drinks per week, which gives approximately 0.40 years of life lost for regular drinking of one alcoholic drink per week, or 2.8 years of life lost for regular drinking of one alcoholic drink per day.
How a “standard” alcoholic drink is defined varies; in the UK it is 8 grams of alcohol and in the US it is 14 grams of alcohol. In addition, most people are not aware of the grams of alcohol in a particular alcoholic drink. So, the accuracy of the reporting of alcohol use is questionable. Alcohol use is also thought to be under-reported in Western settings. If people under-report their alcohol use, then one reported unit of alcohol represents more alcohol than one actual unit, so we are likely overestimating the effect of one actual unit of alcohol.
As such, 2.8 years of life lost per regular drinking of one unit of alcohol per day may be an over-estimate, so I am also providing information about a similar study from China (Millwood IY, Im PK, Bennett D, et al. Alcohol intake and cause-specific mortality: conventional and genetic evidence in a prospective cohort study of 512 000 adults in China1. Lancet Public Health. 2023;8(12):e956-e967). This study is useful in showing clearly that the effect of alcohol is linear (Figure 1). It is also informative because it is different in three important ways:
Women in China rarely use alcohol
Reporting may be more accurate, because only relatively recently have people in China been able to afford alcohol
Alcohol use may not have been lifelong, because the study from China was recruited in 2004-8 with mean age 51 years, giving an average birth year of 1955. Substantial economic development did not start to take place in China until the 1980s
In the study from China, genetically predicted alcohol use had no effect in women because genetics do not represent alcohol use when there is no alcohol use.
Correspondingly, the effects in men are the effects of alcohol not of genetics. The study from China reported a hazard ratio2 for 100g (3.5 oz.) of alcohol/week of 1.07.
Using an actuarial assumption, we can multiply the natural log hazard ratio by -10 which gives the years of life lost. The hazard ratio is 1.07, the natural log of 1.07 is 0.0677, 0.0677 x -10 is -0.677 years. So, this study is estimating 8 months of life lost per 1-2 alcoholic drinks per day for part of people’s lives.
Taken together, we have two estimates, 8 months of life lost per 100g alcohol/week (1-2 drinks per day) for likely part of adult life, and 2.8 years of life lost per one drink per day throughout adult life. The likely answer for regular drinking of one unit of alcohol per day in adulthood is probably somewhere in between these two estimates. The key point is that alcohol use shortens life, with the years of life lost likely proportional to the total amount of alcohol consumed over the lifespan.
Just to contextualize these findings, there are no randomized controlled trials of the effect of alcohol on mortality, which would provide very high quality evidence. Evidence from studies where people are asked to report their alcohol use and then they are followed up to see what happens are difficult to interpret. People who choose to drink alcohol regularly but moderately are probably different from other people, i.e., possibly more moderate in all aspects of their life. So, in these types of study, we do not know if we are seeing the effects of moderate alcohol use or the effects of being a "moderate" person.
I hope this answers your questions, please let me know if you have any further questions
Regards
mary
Round 2: Here’s Dr. Schooling’s emailed answers, sent Feb. 23, to my follow-up questions (in bold below) to her first email:
Has the study been written about by any other media? I haven't seen any other mentions of it, which surprised me given how much attention alcohol consumption has received in recent months.
Mary Schooling: This study has not been written about by the media. We did not send out a press release, so maybe that is our fault. There are a lot of vested interests at stake, which is a bit chilling, because the 2025-30 Dietary Guidelines for America (DGA) [which are the primary source of dietary guidance from the federal government] are under development. The Review of Evidence on Alcohol and Health, generated for the new guidelines, is recommending maintaining the current guidelines3 — two drinks or fewer per day for men, and one drink or fewer per day for women.
There are no trials of the effect of alcohol on mortality. The DGA review excluded Mendelian randomization studies, so it is based on evidence that is unlikely to be valid, as explained here. Mendelian randomization studies are generally less biased than other observational studies.
[Note to readers: Dr. Schooling’s study employed Mendelian randomization aka MR.]
Any idea how the years of lost lifespan might clawed back by those who stop drinking after decades of moderate/social drinking? For example, a healthy man in this 50s who, after drinking regularly since age 21, reduces from 7 drinks a week to only 1 per week? How long, in years, would be required for him to fall out of the category of "7 drinks/week of lost longevity" and fall into the category of "1 drink/week of lost longevity"?
Mary Schooling: That is such a good question about the benefits of giving up alcohol. There is a famous paper on smoking [from 2004] which says that if smokers give up before middle age, then they have the same survival as never smokers. However, smokers have to survive their smoking to become ex-smokers, so this estimate is too optimistic.
My interpretation would be that the years of life lost from alcohol use are proportional to the number of years of alcohol use. So, assuming someone started daily alcohol drinking at age 18 and that life expectancy is 78 years, then giving up alcohol use at age 48 would halve the number of years of life lost for lifetime alcohol drinking from 2.8 years to 1.4 years. And giving up at age 33 would result in only a quarter of the number of years of life lost for lifetime drinking (i.e., 0.7 years). However, we do not have any good evidence on this point.
In retrospect, should your published paper in Nature have been less categorical or specific about equating a certain number of weekly drinks to a rough number of lost years of life?
Mary Schooling: Thank you for asking about the interpretation of our study. We say in the abstract, "Our study indicates that alcohol does not provide any advantages for men or women but could shorten lifespan." I am happy with that. In addition, the estimate of 2.8 years of life lost per one alcoholic drink per day throughout life does have wide confidence intervals from 0.77 years of life lost for one alcoholic drink a day throughout adult life to 4.86 years of life lost. In addition, the estimate we place most emphasis on and give in the abstract (in logged units, apologies for that) is also a conservative estimate for a number of reasons:
Figure 3 bottom right shows an estimate based on 221 SNPs (genetic variants) and an estimate based on only one SNP. The estimate based on only one SNP, is based on a physiologically relevant SNP, so arguably has greater validity. We did not give the larger estimate based on only one SNP in the abstract we gave the more conservation estimate based on 221 SNPs.
The study we used to estimate years of life lost is a study of parental attained age (i.e., age at death or current age of parents, which we used as a proxy for attained age of the participants). This study excluded mothers/fathers who died before 57/46 years to focus on natural deaths. However, alcohol use could result in deaths before these ages due to accidents, but these potentially alcohol related parental deaths were excluded.
Our study uses data on lifespan from the UK Biobank which was recruited at mean age 57 years, so the study is missing people (and the parents thereof) who died at younger ages due to alcohol use. These missing deaths at younger ages contribute more to years of life lost than deaths in the old.
Conversely, as i mentioned before, there is the issue of exactly how much alcohol was reported by the participants to be one alcohol drink, which could have inflated the estimate.
On reflection, i think it would have been better to have presented the results in terms of actual units of alcohol not logged units of alcohol. We also made a lot of assumptions, where we tried to take a balanced approach, as explained above. Finally, the estimates we give are not very precise given the wide confidence intervals and the uncertainty about exactly what an alcoholic drink is.
I hope that helps, please let me know if you have any further questions.
regards
mary
So where does that leave us non-Mendelian randomization experts? How should we functionally think about the imprecise and often vague conclusions about how much lifespan we give by using alcohol?
In a phone interview that preceded her first email response, Dr. Schooling perhaps said it best because she said it simply: “The smaller amount you drink per week, the lesser effect it has,” she told me. “And the larger amount you drink per week, the greater effect it has. What we are clearly saying is that drinking is shortening your life.”
"I have reached the view,” she added, “that alcohol is probably always bad for you.”
The study concluded that, “Among current drinkers, each 100 g [about 3.5 oz.] per week higher alcohol intake was associated with higher mortality risks from cancers, cardiovascular disease, liver diseases, non-medical causes, and all causes.”
A hazard ratio is a statistical measure that compares the rate at which an event (like death or disease progression) occurs in one group compared to another group over a period of time, essentially indicating the relative risk of an event happening between the two groups. A hazard ratio of 1, for instance, means the event occurs at the same rate in both groups, while a ratio greater than 1 indicates a higher risk in the first group and a ratio less than 1 indicates a lower risk.
Here’s my take on the National Academy of Sciences’ unscientific recommendation that 2 drinks/day for men, and 1 drink/day for women, promote good health.
Damn. By these metrics I was destined to live to at least 200 years old but I foolishly smoked from age 18 to 30, as well as having at least two drinks per week on average. For the following 15 years I would have had on average 7-10 drinks per week. Then, as best I can recall, I drank on average six bottles of wine (mostly reds) plus an average 7 nips of single malt scotch each week from age 45 to age 65 when I lost the taste for alcohol and now only drink at Christmas and a couple of other occasions each year. It’s beyond me to calculate the years ‘lost’ (maybe a side effect?) but as I head towards 72 in reasonable health for my age I’m not sure I’d want to add another 100+ years to whatever the expiration date on my birth certificate is. But thanks for explaining the research.
Sobering…