How much alcohol is too much?

What 107 studies found about drinking and the risk of dying, and what health agencies advise now. Every number links to its source.

The short version

Once the studies' flaws were corrected, light drinkers didn't clearly live longer than people who never drank.
The risk of dying was clearly higher above about 3 drinks a day, and for women from about 1.8.
Alcohol causes at least seven kinds of cancer, and WHO Europe says no level of drinking is safe for health.
Canada calls 1 or 2 drinks a week low risk, the UK advises no more than 14 units a week, and the US says to drink less, and that some people shouldn't drink at all.

Rate your own drinking

Count the drinks in a typical week, then find your row. A US standard drink holds about 14 g of alcohol: a 12-ounce can of 5% beer, a 5-ounce glass of 12% wine or a 1.5-ounce shot of 40% spirits (NIAAA). A 175 ml glass of 12% wine counts as about 1.2, and a pint of 5.2% beer as about 1.7 (NHS).

US drinks a week	Canada's rating	Risk of dying in 107 studies, vs never drinking
0	Not drinking has benefits	The comparison group
1 to 2	Low risk	No clear difference
3 to 6	Moderate risk	No clear difference
7 to 12	Increasingly high risk	No clear difference
13 to 22	Increasingly high risk	21% higher for women; for everyone, 5% higher, which could be chance
23 to 32	Increasingly high risk	19% higher (34% for women)
33 or more	Increasingly high risk	35% higher (61% for women)

Canada's ratings come from its 2023 guidance (CCSA, 2023), which counts in drinks of 13.45 g, a little smaller than US ones (CCSA final report). The UK's advice of no more than 14 units a week is about 8 US drinks. The study's rows are its daily drinking levels times seven, and "no clear difference" means the gap could be chance in either direction.

What 107 studies found

A 2023 analysis pooled 107 studies that tracked 4,838,825 people for years, published from 1980 to 2021, with 425,564 deaths from any cause. It compared drinkers with people who never drank (Zhao, 2023).

Drinks a day vs risk of dying (1.0 = never drank). Only the two right bars clearly differ.

In the studies as they stood, light drinkers had a 15% lower risk of dying than the non-drinkers. After the authors corrected for ex-drinkers among the non-drinkers and other differences between studies, the gap was 7%, and it could be chance. At each drinking level, the corrected risk of dying compared with people who never drank was:

Very occasional drinking, under two-thirds of a drink a week: 4% lower, which could be chance.
About 0.1 to 1.8 drinks a day: 7% lower, which could be chance.
About 1.8 to 3.2 drinks a day: 5% higher, which could be chance.
About 3.2 to 4.6 drinks a day: 19% higher.
4.6 drinks a day or more: 35% higher.

The paper counts alcohol in grams. Here a drink is 14 g, a US standard drink, so the levels above are 1.3 to under 25 g a day, 25 to under 45 g, 45 to under 65 g, and 65 g or more, and the drink ranges are rounded.

One reason light drinking looked healthy

In 86 of the 107 studies, the non-drinkers included ex-drinkers or occasional drinkers. Many people stop drinking for health reasons, which can make non-drinkers look sicker and light drinkers healthier next to them. Ex-drinkers had a 26% higher risk of dying than people who never drank.

Women and men

Below about 1.8 drinks a day, women's and men's risks, each compared with people of their own sex who never drank, were similar. Above that, the risk was higher sooner for women:

Women drinking about 1.8 to 3.2 drinks a day had a 21% higher risk of dying than women who never drank, rising to 34% at 3.2 to 4.6 drinks and 61% at 4.6 or more.
For men, the risk was clearly higher from about 3.2 drinks a day: 15% higher at 3.2 to 4.6 drinks, and 34% at 4.6 or more.

Cancer

The International Agency for Research on Cancer lists alcohol among the substances proven to cause cancer in people, and alcohol causes at least seven kinds, including bowel cancer and breast cancer in women. WHO Europe says no level of drinking is safe for health, because no amount has been found below which alcohol's cancer-causing effects stop. It adds that around half of the cancers caused by alcohol in the WHO European Region occur in people who aren't heavy drinkers (WHO Europe, 2023).

What health agencies advise

Canada, 2023. Not drinking has benefits. 1 or 2 drinks a week is low risk, 3 to 6 is moderate risk, and 7 or more is an increasingly high risk (CCSA, 2023).
UK. Men and women are advised not to drink more than 14 units a week on a regular basis. A unit is 8 g of alcohol, so 14 units is about 8 US drinks a week (NHS).
US, 2025 to 2030. The dietary guidelines give no daily number. They say to "consume less alcohol for better overall health", and that some people should avoid alcohol completely, including pregnant women, people recovering from alcohol use disorder, and people on medicines that interact with it (Dietary Guidelines for Americans, 2025 to 2030).

How far to trust these numbers

These studies tracked people, which can show links but can't prove that drinking caused the deaths.
Most measured drinking by asking people, often only once, and people tend to under-report it. Even people who say they never drank are sometimes misclassified.
The corrections were made with statistics across whole studies. They can reduce the flaws but not remove them.
The drinking levels are averages, and the cut-offs between them are the authors' categories. Risk doesn't jump at them.
The analysis counted deaths from any cause. It doesn't separate out cancer, liver disease or injuries.