Ask a public health agency how much alcohol is low-risk and the answer arrives as a number: in the United States, no more than two drinks a day for men and one for women. Where that number comes from is another matter. Low-risk drinking limits are not measurements of a biological safe line; they are written-down judgments about how much lifetime risk a person should accept, translated into drinks per week.
A modelling study published on 19 October in the International Journal of Epidemiology makes that machinery unusually visible. Commissioned to inform an update of Aotearoa New Zealand’s low-risk drinking guidelines, it estimates how many standard drinks a week a person can average before crossing two conventional risk thresholds. New Zealand’s current advice, last revised in 2011, sits at 15 standard drinks a week for men and 10 for women.
Using the World Health Organization’s relative-risk functions, the authors found that men cross a 1-in-100 lifetime risk of alcohol-attributable death at 8.6 standard drinks a week, with a 95% uncertainty interval of 3.1 to 12.0. For women the figure was 9.2 drinks, interval 3.7 to 12.4. A standard drink here is 10 grams of ethanol, smaller than the 14-gram US standard drink described by MedlinePlus.
What the model was actually asked to do
The New Zealand Alcohol Intervention Model is a multi-state life-table simulation, not a survey. It ran four cohorts of 10,000 twelve-year-olds — Māori and non-Māori, male and female — across their modelled lifetimes, with disease and injury rates fixed at 2018 levels, and compared drinking scenarios from 0 to 100 grams of ethanol a day against lifetime abstention. The output was years of life lost, with one alcohol-attributable death proxied at 17.5 years of life lost.
Two thresholds were reported. The 1-in-100 line, the benchmark most jurisdictions including New Zealand have used, corresponds to 175 years of life lost per 1,000 people; the stricter 1-in-1,000 line to 17.5.
The choice of risk functions alone shifted the result. With the Global Burden of Disease Study’s 2025 library instead of WHO’s, the 1-in-100 threshold moved to 12.0 drinks a week for men and 10.9 for women, and at the stricter line the WHO analysis crossed at 3.0 drinks a week for men and 6.1 for women. In grams, the headline figures come to 86 grams a week for men and 92 for women, against 81 in Canada’s 2023 guidance, 100 in Australia’s 2020 update and 112 in the UK’s 2016 advice.
Three decisions that move the number
The study’s most useful contribution is not the final figure but the demonstration that a guideline number is assembled from choices, each of which can swing the answer by several drinks a week.
The first is which health outcomes count. The GBD grades each risk-outcome pairing on a one-to-five-star evidence scale, and removing the conditions rated one star — coronary heart disease, intracerebral haemorrhage, ischaemic stroke and stomach cancer — pulled the 1-in-100 threshold down to 2.7 drinks a week for men (interval 2.1 to 4.6) and 3.5 for women (interval 2.9 to 4.4). At the stricter line it fell to one drink a week or less for both sexes. The authors tie that to low-dose cardioprotective effects disappearing once between-study heterogeneity is accounted for.
The second is drinking pattern. The model adjusts relative risks for binge drinking at each average intake level, because injury risk in the hours after heavy drinking is not captured by a weekly average. Strip that adjustment out and the 1-in-100 threshold rises to 12.2 drinks a week for men and 10.3 for women under the WHO functions, and to 18.5 and 12.8 under the GBD ones. In the New Zealand data, 17% of respondents who averaged less than 2 grams a day were classified as binge drinkers.
The third is the threshold itself, and here the paper turns openly normative. A 1-in-100 lifetime risk is often justified on the grounds that drinking is a voluntary exposure. The authors push back: much alcohol harm falls on people who did not choose to drink, through injuries to others and through foetal alcohol spectrum disorder, and the drinker’s own choice is questionable once dependence is in the picture. They conclude New Zealand’s guidelines are outdated.
A second kind of evidence: what a blood test finds
If the modelling describes a population, a Swedish cohort published three days earlier in Acta Orthopaedica describes a hospital ward. Clinicians at Skåne University Hospital in Malmö added phosphatidylethanol, or PEth, to the routine admission blood panel for patients aged 65 and older arriving with a fall-related fracture, and measured it between February 2023 and August 2024.
PEth forms only when alcohol is present. It takes more than a week of regular drinking above moderate amounts to accumulate, and its half-life runs from four to ten days, so it reports sustained drinking rather than last night’s glass.
Of the 505 patients included, 13% had PEth in the hazardous range of 0.3 micromoles per litre or above, 8.9% were in the intermediate band, and 78% were low or abstinent. Continuous consumption — anything at or above 0.05 micromoles per litre — was present in 19% of the women and 29% of the men. Among non-frail patients, 18% were in the hazardous range and 11% in the intermediate one; non-frail patients drank more than frail ones.
In a propensity-matched analysis, sustained drinking was associated with higher odds of low sodium on admission, an odds ratio of 2.3 with a confidence interval of 1.1 to 4.5. A trend toward lower 90-day mortality in that group, odds ratio 0.2 with a confidence interval of 0.1 to 1.0, sits on the edge of significance and is exploratory. The authors’ summary is deliberately flat: at least one in five older patients with fall-related fractures has continuous alcohol consumption, and otherwise no apparent negative effects were identified in the short term.
The quietest number may be the one about who was not tested: of 1,094 eligible patients, 589 were excluded because no PEth result existed. A marker that is missing for more than half of eligible patients does not yet function as routine screening.
What the existing guidance says
US agencies already say less is better without naming a threshold below which risk vanishes. The Centers for Disease Control and Prevention sets moderate drinking at two drinks or less a day for men and one or less for women, binge drinking at four or more drinks for women or five or more for men on an occasion, and heavy drinking at eight or more drinks a week for women or 15 or more for men. Its summary line is that drinking less is better for your health than drinking more.
CDC figures put about 178,000 US deaths a year in 2020–2021 in the alcohol-attributable column, a 29% increase from an estimated 138,000 in 2016–2017. Roughly two-thirds came from chronic conditions and about 61,000 from binge drinking or a single heavy occasion, including crashes, overdoses and suicide. In the agency’s surveillance data, 17% of adults report binge drinking and 6% report heavy drinking.
MedlinePlus adds the caveat underneath all of this: even though moderate drinking may be safe for many people, there are still risks, and moderate drinking can raise the risk of death from certain cancers and heart diseases. It lists people who should not drink at all, among them anyone in recovery from an alcohol use disorder, anyone unable to control how much they drink, people under 21 or pregnant, and those taking medicines that interact with alcohol.
The cancer signal travels least well into everyday conversation. The National Cancer Institute, quoted on the NIAAA’s page on alcohol’s effects on the body, states that even people who have no more than one drink a day, along with people who binge drink, have a modestly increased risk of some cancers. The World Health Organization calls alcohol an established carcinogen and attributes 4.4% of cancers diagnosed globally in 2019, and 401,000 cancer deaths that year, to alcohol. Its fact sheet counts 2.6 million alcohol-attributable deaths worldwide in 2019, two million of them among men, and about 400 million people aged 15 and over living with alcohol use disorders.
None of that makes a weekly total a medical instrument. If you are weighing your own drinking, our guide to cutting 500 calories a day covers the calorie side, and our piece on drinks that fit a weight-loss plan covers what to drink instead. Neither substitutes for advice about your own health.
What we did not find
We found no randomised trial that assigns people to different drinking levels for years and follows disease outcomes; that study cannot ethically exist, which is why this literature is built from modelling and cohorts. The New Zealand work is a simulation whose answer depends on which risk functions and which outcomes are included, and its authors say so. The Swedish cohort is single-centre, observational, followed patients for 30 days for complications and 90 days for death, and had no PEth result for more than half its eligible patients.
We also found no trial showing that a defined period of abstinence, such as a dry month, changes long-term disease risk. The modelling suggests risk rises roughly in step with average intake, without a plateau at low levels, which is a statement about populations over lifetimes, not a promise that a break resets anything.
Nothing here is a diagnosis, a prescription, or a recommendation for an individual reader. Guidelines describe populations; your own situation depends on your history, your medications, whether you are pregnant, and conditions such as liver disease. If you drink and want to know whether it is affecting your health, or if cutting down is harder than you expected, that is a conversation with a healthcare professional, and our medical disclaimer explains where this publication’s limits sit.