The most comprehensive analysis of global alcohol data ever assembled found that the amount of alcohol associated with the lowest overall risk to health is zero drinks a week — not because every sip is equally dangerous, but because even the modest benefit some studies find for a couple of conditions at low doses is outweighed, once every health outcome is modeled together, by alcohol's effect on cancer risk and injury. The World Health Organization reached a similar conclusion in 2023, stating plainly that no level of alcohol consumption has been shown to be safe.
In brief:
- A 2018 analysis spanning 195 countries modeled alcohol's effect on 23 separate health outcomes together and found the risk-minimizing amount was zero drinks per week (95% uncertainty interval 0.0–0.8).
- Among adults 15 to 49, alcohol was the leading risk factor for death and disability worldwide in 2016 — driven mostly by injuries and communicable disease, not the liver damage most people associate with heavy drinking.
- Among adults 50 and older, cancer is the bigger story: it accounted for more than a quarter of alcohol-attributable deaths in women and close to a fifth in men in that age group.
Why did researchers conclude the safest amount is zero?
Alcohol's relationship with health isn't one-directional. Some studies have found small amounts associated with modestly lower risk of a few specific conditions, like ischemic heart disease — while the same small amounts raise the risk of several cancers, tuberculosis, and injuries, even at low doses. The question that actually matters for a health recommendation isn't "does any single outcome improve at low doses," it's "what happens to your overall risk across everything alcohol touches."
That's the question the GBD 2016 Alcohol Collaborators set out to answer. The group combined 694 data sources on how much people actually drink with 592 studies linking alcohol to 23 different health outcomes, across 195 countries and territories, from 1990 to 2016.¹ When they modeled the net effect of different consumption levels across all 23 outcomes combined — rather than picking any one disease in isolation — the level associated with the lowest total risk of death or disability came out to zero drinks per week, with a 95% uncertainty interval of 0.0 to 0.8 standard drinks. In plain terms: the model's best estimate was zero, and even the upper edge of its uncertainty allowed for well under one drink a week.
What is alcohol's biggest risk, by age?
The same analysis found the picture changes substantially depending on how old you are.¹ Among adults 15 to 49, alcohol was the single leading risk factor for death and disability of any kind, worldwide, in 2016 — responsible for an estimated 3.8% of female deaths and 12.2% of male deaths in that age band. The driver in that age group isn't primarily liver disease; it's tuberculosis, other communicable diseases, and injuries such as road traffic crashes and self-harm, categories that don't typically come to mind when people picture "alcohol harm."
Among adults 50 and older, the leading driver shifts to cancer: the analysis attributed 27.1% of alcohol-related deaths in women and 18.9% in men in that age group to cancer specifically. Overall, across all ages, alcohol ranked as the seventh leading risk factor for combined deaths and disability worldwide in 2016, accounting for an estimated 2.2% of age-standardized female deaths and 6.8% of male deaths.
Has the position changed since 2018?
If anything, it's hardened. In January 2023, the World Health Organization issued a statement built on this same body of evidence, concluding that no level of alcohol consumption has been shown to be safe for health.² The reasoning centers on cancer specifically: alcohol is classified as a Group 1 carcinogen, and current evidence doesn't identify a threshold below which its cancer-causing effect switches off. The statement also addressed the heart-health question directly — any possible cardiovascular benefit some studies associate with light drinking does not, in WHO's assessment, outweigh the cancer risk at an individual level.
What the research says
The GBD 2016 Alcohol Collaborators, drawing on 694 data sources covering how people actually drink and 592 studies linking alcohol to 23 health outcomes across 195 countries from 1990 to 2016, modeled the combined effect of different consumption levels across all of those outcomes together and found the risk-minimizing level was zero drinks per week, with a 95% uncertainty interval of 0.0 to 0.8 standard drinks.¹ The limit: this is a population-level statistical model that synthesizes many underlying studies — most of them observational — so it describes the net pattern across large populations rather than predicting any individual's personal risk, and the width of the uncertainty interval itself is a signal of how much the estimate could shift as more data comes in.
The World Health Organization's 2023 statement draws on this same evidence base specifically for cancer risk, concluding that current research doesn't identify a consumption threshold below which alcohol's carcinogenic effect doesn't apply.² That statement is a position summary issued by a global health body synthesizing existing research, not a new study generating its own data, so it carries the same observational-evidence limits as the research underlying it — and like the 2018 analysis, it describes a population-level finding, not a guarantee about what happens to any one person who drinks a given amount.
What you can do this week
- Think in standard drinks, not "glasses." A standard drink is about 14 grams of pure alcohol — roughly 12 oz of regular beer, 5 oz of wine, or 1.5 oz of distilled spirits — and pour sizes at home or at a restaurant often run larger than that without looking different.
- Treat any amount as a risk you're weighing, not a "safe baseline" to hit. The research above describes population-level risk, not a personal floor to drink up to; if you choose to drink, there's no dose the evidence identifies as risk-free.
- Bring your actual drinking pattern to your next checkup. How much and how often matters more than how it feels in the moment, and it's a normal, useful thing to discuss with whoever manages your general health — not something to downplay or round down.
The research doesn't say alcohol is uniquely catastrophic at any amount, or that moderate drinkers are courting disaster. It says the number researchers keep landing on, once they account for every health outcome alcohol affects rather than just the ones that get discussed most, is zero — and that number has only become more settled, not less, as the evidence has grown.
