A couple of drinks a night
Peer-reviewed research · 3 sources · Last verified 2026-08-24
A couple of drinks most evenings is the level almost nobody worries about. It is also the level where the liver evidence stops being reassuring. Pooled research links that intake to a markedly higher rate of serious liver disease, measured against people who never drank at all - and, in men, to more new high blood pressure. What barely moves at this level is the overall death rate. That is the honest half: the damage shows in the organ long before it shows in the statistics about dying.
The one honest fix
When the glass is empty, put the bottle down and decide again before pouring. Let the second drink be something you choose, not something that just follows the first.
Free · behavioral
What the evidence says
Drinking 25 g of alcohol a day - roughly two standard drinks - was associated with 1.81 times the rate of developing liver cirrhosis compared with lifetime abstention, and the pooled curve rises steeply above that level
+81% · RR 1.81, 95% CI 1.68-1.94
Limitations & how to read this number
Two things about this figure are worth holding on to. The first is what it is compared against. The reference group here is lifetime abstention - people who never drank - and the reviewers deliberately harmonised the included studies so that this, rather than a vague 'non-drinker' category, was the comparison. That matters in alcohol research more than almost anywhere else, because a group labelled non-drinkers usually contains people who used to drink and stopped, often because they had already become unwell. A clean comparison group makes this estimate more trustworthy than most in this field, and it is worth saying so. The second is what it is not. It is a relative rate, not a personal forecast: cirrhosis is uncommon in the general population, so a rate that is roughly four fifths higher is still a small absolute number for most people, and the great majority of people drinking at this level will never develop it. The pooled figure covers both sexes together. The reviewers noted that in sex-specific work women sit on a steeper curve at any given intake, but they did not publish separate figures for men and women here, so nothing on this page should be read as the same number for everyone. The evidence is observational. Of the 44 studies, 26 followed people forward in time and 18 were case-control studies, which look backwards and cannot establish which came first. Alcohol intake was almost always self-reported at a single point in time, and self-reported drinking is usually an underestimate. Finally, the participant and case counts for this review - 5,122,534 people and 15,150 cases of cirrhosis - cover the whole review, both the illness analysis and the death analysis, and no headcount is reported for the estimate quoted here on its own.
SourceLlamosas-Falcón L et al. (2024) · Hepatology International · 10.1007/s12072-023-10584-z
Men drinking 1 to 2 drinks a day had 1.19 times the rate of developing high blood pressure compared with abstainers (95% CI 1.07 to 1.31), while women drinking the same amount showed no increased rate (0.94, 95% CI 0.88 to 1.01)
+19% · RR 1.19, 95% CI 1.07-1.31
Limitations & how to read this number
This one is not the same for everybody, and that is the finding rather than a footnote. In this review a standard drink is 12 g of pure alcohol, so 1 to 2 drinks a day is the band a couple of drinks falls into. At that intake men had a higher rate of developing high blood pressure than abstainers; women at the same intake did not, and their estimate sits right on the line of no difference. Women's rates did rise at higher intake. The reviewers also reported that they found no sign of a protective effect of drinking in women at any level, so the flat result at 1 to 2 drinks a day should be read as nothing gained rather than as a benefit. The studies disagreed with one another quite a lot - moderately so among men, more so among women - which widens the honest uncertainty around both figures. Everything here is observational: nobody was assigned to drink or not drink, and people who drink differ from people who do not in weight, salt, stress, sleep, smoking and income, only some of which the included studies could account for. All the drinking measures were self-reported. The review's totals - 361,254 participants and 90,160 new cases of high blood pressure, of whom 125,907 participants and 32,426 cases were men - describe the whole body of studies, not the particular estimate quoted here, so no headcount is printed beside it.
SourceRoerecke M et al. (2018) · Journal of the American Heart Association · 10.1161/JAHA.117.008202
Drinkers consuming 25 to under 45 g of alcohol a day showed no clear difference in overall death rate from lifetime non-drinkers, once the analysis corrected for former-drinker and occasional-drinker bias (1.05, 95% CI 0.96 to 1.14, from 80 studies) - although women at that intake had a higher rate (1.21, 1.08 to 1.36) while men did not (1.01, 0.93 to 1.10)
no significant difference · RR 1.05, 95% CI 0.96-1.14
Limitations & how to read this number
This is a null, and a null has two halves that both have to be said out loud. It does not show that drinking at this level is safe, and it does not show that it is harmful: the confidence interval runs from slightly below no difference to slightly above it, so overall death rates in this pooled analysis could not tell the two apart. What it is useful for is timing. Damage to an organ shows up in the organ long before it shows up in a death rate, and this is the number that keeps the rest of the page honest - a couple of drinks a night is associated with more liver disease and, in men, more high blood pressure, but at this level the evidence does not show more people dying. The split by sex is part of the finding and is printed above rather than tucked away: women at this intake had a higher overall death rate than lifetime non-drinkers while men did not, which is the same direction the liver reviewers described. The comparison group is the reason this review exists. In older research the people labelled abstainers often included former drinkers, many of whom had stopped because they were already ill, and occasional drinkers who were not abstainers at all - a comparison that quietly flatters anyone still drinking. This analysis modelled those biases out and compares against lifetime non-drinkers instead. Everything here remains observational, the studies disagreed with each other substantially, and the review's totals of 4,838,825 participants and 425,564 deaths belong to all 107 studies together, not to the estimate quoted here.
SourceZhao J et al. (2023) · JAMA Network Open · 10.1001/jamanetworkopen.2023.6185
How to actually do it
Interrupt the pour that nobody decides on. When the first glass is empty, put the bottle down and decide again - out loud if it helps. Sometimes the answer will be yes, and this page does not argue with that. But a drink that comes from a decision is a different thing from one that simply follows the last one, and the automatic ones are the easiest to let go. This costs nothing and is not a reason to buy a low-alcohol anything. If putting the bottle down turns out to be hard, that is information worth taking to a doctor rather than a signal to try harder.
Supporting findings (2)
The same liver review reported that at 25 g of alcohol a day the rate of death from liver cirrhosis was 2.65 times that of lifetime abstainers (95% CI 2.22 to 3.16), a steeper gradient than for developing the disease
+165% · RR 2.65, 95% CI 2.22-3.16
This figure is kept off the headline on purpose, and it is worth explaining why rather than simply omitting it. It describes deaths from cirrhosis rather than diagnoses of it, and death from cirrhosis is a much rarer event, so a larger relative figure is sitting on top of a much smaller base rate. It also tends to be recorded in people whose liver disease was already advanced, and heavier or longer drinking histories are concentrated there, which is part of why the gradient is steeper than for the illness itself. The same caveats as the main liver figure apply: both sexes are pooled, the studies are a mixture of forward-looking and backward-looking designs, drinking was self-reported, and the review's participant and case totals cover all 44 studies rather than this estimate alone. It is included here because leaving out the larger number from the same table would be a kind of editing this page does not do.
SourceLlamosas-Falcón L et al. (2024) · Hepatology International · 10.1007/s12072-023-10584-z
In the same mortality review, the apparent survival advantage of light drinking did not hold up once the analysis corrected for the fact that the comparison group of abstainers contained former drinkers and occasional drinkers, leaving low-volume drinkers - 1.3 to under 25 g of alcohol a day - at a relative risk of 0.93 against lifetime non-drinkers (95% CI 0.85 to 1.01), an interval that includes no difference at all
no significant difference · RR 0.93, 95% CI 0.85-1.01
This figure belongs to a smaller amount of drinking than the rest of this page - up to about two drinks a day at the very top of the band, and typically much less - and it is here for one reason: it is the number that answers the belief most people are carrying. For decades the research seemed to show that people who drank a little outlived people who did not drink at all. When the comparison group is repaired, so that drinkers are compared with people who never drank rather than with a group that quietly contained people who had stopped drinking because they were ill, the advantage fades to nothing measurable. Read it as the collapse of a claimed benefit, not as a verdict on a particular glass, and not as evidence of harm at that level either - the interval crosses no difference. It is observational, pooled across studies that disagreed with one another substantially, and no participant count is reported for this particular estimate.
SourceZhao J et al. (2023) · JAMA Network Open · 10.1001/jamanetworkopen.2023.6185
Go deeper
Open what you care about
What "a couple of drinks" actually is, in grams
Almost every number in this literature is expressed in grams of pure alcohol per day, because a "drink" means different things in different countries and different things in different glasses. A UK unit is 8 g. A standard drink is 12 g in the hypertension review quoted on this page, and 14 g in the United States. A 175 ml glass of wine at 13% is about 18 g on its own; a 500 ml bottle of 5% beer is about 20 g. So "a couple of drinks" is not a fixed quantity - depending on what is in the glass it lands somewhere between roughly 20 and 35 g a day. The liver estimate on this page is anchored at 25 g a day, which sits squarely in that range, and the mortality estimate covers the band from 25 to just under 45 g a day. This is worth pausing on, because the most common way people underestimate their own intake is not lying about how many drinks they had. It is pouring a large glass and counting it as one. This is also why the page prints no number of alcohol-free days. The advice most people have absorbed is two or three a week; we went looking for where that came from and could not find it in the research. No trial and no meta-analysis tests one number of drink-free days against another. The UK Chief Medical Officers' guideline, the most-quoted official advice in English, sets a weekly limit and says only that someone wanting to cut down should try to have several drink-free days - several, with no number attached. The best-known numbered version came from a public campaign run jointly with a charity funded by the alcohol industry, and the peer-reviewed work that campaign generated examines the partnership rather than whether the advice worked. What the evidence does support is the shape of the curve: risk rises with the dose, and it rises faster the further along you go.
The liver figure, and why its comparison group is unusually clean
The anchor for this page is a 2024 systematic review and dose-response meta-analysis in Hepatology International that pooled 44 studies of alcohol and liver cirrhosis, covering 5,122,534 participants and 15,150 cases across both the illness and the death analyses. At 25 g of alcohol a day, the pooled relative risk for developing cirrhosis was 1.81 (95% CI 1.68 to 1.94) compared with lifetime abstention. The curve does not continue in a straight line: at 50 g a day the same model gives 3.54 (95% CI 3.29 to 3.81), and at 100 g a day, 8.15 (95% CI 7.46 to 8.91). Doubling the dose does considerably more than double the figure. The phrase "compared to lifetime abstention" is doing quiet but important work. The reviewers harmonised the included studies so that lifetime abstainers were the reference group, and in alcohol research that is not a formality - it is the single correction that has overturned more findings in this field than any new dataset. A group described simply as "non-drinkers" tends to contain people who stopped drinking, often because they had become ill, which makes anyone still drinking look healthy by comparison. Two honest caveats sit alongside the figure. The estimate pools men and women together; the authors note that in sex-specific work women sit on a steeper curve at any given intake, but they did not publish separate numbers here, so we do not print any. And 18 of the 44 studies were case-control studies, which look backwards from the diagnosis and cannot establish which came first.
Blood pressure, where the answer is genuinely different for men and women
The second source is a 2018 systematic review and meta-analysis in the Journal of the American Heart Association, pooling 20 reports from 18 cohort studies with 361,254 participants and 90,160 new cases of high blood pressure. It is the cleanest design in this module's source set: cohort studies only, everybody free of hypertension at the start, followed forward. Its result splits by sex and refuses to be simplified. Among men, drinking 1 to 2 drinks a day - a standard drink being 12 g of alcohol in this review - was associated with a relative risk of 1.19 (95% CI 1.07 to 1.31) for developing high blood pressure compared with abstainers. Among women at the same intake, the relative risk was 0.94 (95% CI 0.88 to 1.01), which is no measurable difference. Women's risk did rise at higher intake, at 1.42 (95% CI 1.22 to 1.66) from three drinks a day upwards. The reviewers add one line that is easy to miss and worth repeating: they found no evidence of a protective effect of drinking in women at any level. So the flat result for women at a couple of drinks a day means nothing gained, not something gained. Both estimates carry substantial disagreement between studies, more so among women than men, and all of it is observational and self-reported.
The part most alcohol articles leave out - the death rate barely moves here
If you have read anything about alcohol in the last few years, you have probably absorbed the idea that every drink is shortening your life. The largest pooled analysis of the question does not support that at the dose this page is about. A 2023 systematic review and meta-analysis in JAMA Network Open pooled 107 cohort studies with 4,838,825 participants and 425,564 deaths, and rebuilt the comparison group so that drinkers were measured against lifetime non-drinkers rather than against a mixed bag of former and occasional drinkers. In the fully adjusted model, people drinking 25 to under 45 g a day had a relative risk of death of 1.05 (95% CI 0.96 to 1.14), drawn from 80 studies and 146 risk estimates - an interval that includes no difference. The increases the review does find are higher up: 1.19 (95% CI 1.07 to 1.32) at 45 to 64 g a day and 1.35 (95% CI 1.23 to 1.47) at 65 g or more. The same review splits that middle band by sex, and we print both halves because printing one would be misleading: women at that intake had a relative risk of 1.21 (95% CI 1.08 to 1.36) and men 1.01 (95% CI 0.93 to 1.10). Two things follow. Reading this page as "two drinks a night will kill you" would be going beyond the evidence. And a null in the death rate is not a clean bill of health - it is the ordinary sequence of chronic illness, in which an organ is affected years before mortality statistics notice. One disclosure: the authors of that review are based at a Canadian substance-use research institute and are well-known critics of the older protective-effect literature, as their opponents are well-known defenders of it. We quote a preregistered meta-analysis of 107 cohorts rather than an opinion, but the field is contested in both directions and you should know that.
Who this page is not for
This is written for someone whose drinking is a routine they have never examined. It is not written for, and is not adequate for, anyone whose relationship with alcohol is harder than that. If you find you cannot easily leave a drink alone, if you drink to steady yourself or to get to sleep, if you have tried to cut down and could not, or if a day or two without a drink brings shaking, sweating, nausea or bad anxiety, that is a medical matter and not a question of willpower. Stopping suddenly after heavy or dependent drinking can be genuinely dangerous and should be done with a doctor, not alone and not on the strength of an article. Nothing on this page diagnoses anything, nothing here is treatment, and none of the research quoted here studied people with alcohol dependence. The same applies if you already have liver disease, take medication that interacts with alcohol, are pregnant or trying to conceive, or have been given specific advice by a clinician: their advice governs and this page does not. Free, confidential help exists in most countries through national alcohol helplines and through any GP, and asking about it is an ordinary thing to do.
FAQ
What about cancer? I thought that was the main risk.
It is a real part of the picture, and it is deliberately not the focus of this page. Alcoholic beverages, and the acetaldehyde the body forms when it breaks alcohol down, are classified by the International Agency for Research on Cancer in Group 1, its highest category of evidence. That classification describes how confident the evidence is that a link exists - it says nothing about how large the risk is, which is a separate question answered by different studies. We have written about that distinction between hazard and risk elsewhere on this site. We have not printed a cancer number here because the sources this page is built on are about the liver, blood pressure and overall mortality, and quoting a figure from a study we have not put through our own verification would break the rule this site runs on. If you want the short version: the cancer association is well established, it rises with dose like everything else here, and it is not the reason this page exists.
I find it hard to skip the second drink. Does that mean something?
It might, and it is worth taking seriously rather than treating as a character flaw. Difficulty leaving alcohol alone, drinking to steady yourself or to get to sleep, repeated attempts to cut down that did not hold, or feeling shaky, sweaty, nauseous or very anxious after a day or two without a drink are all reasons to speak to a doctor. That last group especially: withdrawal after heavy or dependent drinking can be medically dangerous, and stopping suddenly and alone is not the right move - it is one of the few situations where cutting down should be supervised. Nothing on this page is treatment, none of the research quoted here studied people with alcohol dependence, and none of it is a way to work out whether you have it. Your GP has had this conversation many times, most countries have a free confidential alcohol helpline, and asking is an ordinary thing to do rather than an admission of anything.
Sources
- Llamosas-Falcón L et al. (2024). Hepatology International. 10.1007/s12072-023-10584-z
- Roerecke M et al. (2018). Journal of the American Heart Association. 10.1161/JAHA.117.008202
- Zhao J et al. (2023). JAMA Network Open. 10.1001/jamanetworkopen.2023.6185
Last verified 2026-08-24 · adversarial fact-check + founder science gate.
Corrections are published openly. Educational content, not medical advice — talk to a professional about your own situation.