The question sounds straightforward. One number, one threshold, one line you stay under. Seven studies across four outcomes, and more than 1.6 million people in the weight studies alone, produced an answer that is not a number. It is a map of what was measured, and at what dose.
Ask how many drinks is too many and you will meet a different number everywhere you look. The studies behind this page do not give one either: each measured a different outcome, in a different way, at different doses.
Sleep is the one outcome here with a dose model. When Gardiner and colleagues pooled 27 studies and ran a formal dose-response model, REM sleep started on average 18 minutes later after alcohol than without it in the 14 studies that measured it, across the doses they gave. In their model the delay became significant from about 0.35 g of alcohol per kg of body weight, which they put at about two standard drinks of 10 g for a 62 kg adult. How people felt about their sleep was not pooled.
Fat burning was measured at one dose in each of two small studies. In one, six men lying in hospital beds were given alcohol through a vein (0.75 g per kg of body weight), and their fat burning fell 87% within the hour. In the other, eight men drank two vodka drinks (24 g of alcohol), and their fat burning was 73% lower one to two hours later. Neither study tested fat burning at a smaller dose, so neither shows the dose at which it starts to drop.
Neither followed fat burning until it recovered, so neither shows how long the drop lasts. Very little of the alcohol became fat: in the second study the liver made at most about 0.8 g of new fatty acids in the six hours after the two drinks, an upper-bound estimate. Its authors conclude that the liver shifts the body’s fat balance by supplying acetate as fuel, not by turning alcohol into fat.
Weight was studied by intake band, mostly in snapshot studies. In a review of 127 studies of more than 1.6 million people, cited here from its preprint (not peer reviewed), most were cross-sectional snapshots, and its dose analysis of those found that moderate drinkers, 14 to 28 g of alcohol a day, showed no significant link with being overweight and had 16% lower odds of obesity than non- or light drinkers. Heavy drinkers, over 28 g a day (nearly three drinks of 10 g), had 40% higher odds of overweight or obesity than non- or light drinkers, though not of obesity on its own. Snapshot studies cannot show cause, and, pooled, the studies that followed people over time found no significant link.
Food intake and testosterone were measured too; they come below.
And muscle protein synthesis? In the research Parr and colleagues published, eight active young men drank about 12 standard drinks over three hours after a combined workout. Measured two to eight hours after the workout, muscle protein building was 24% lower than with protein alone when they also had protein, and 37% lower when they had carbohydrate instead. None of the studies behind this page measured what two or three drinks do to muscle protein building.
So the answer is not one threshold but four outcomes, tested in four different ways: sleep across many doses, weight by intake band, and fat burning and muscle at a single dose per study.
Why the Fat-Burning Drop and the Weight Data Don’t Meet
If alcohol cuts fat burning this sharply, why were moderate drinkers not heavier in the weight studies?
These studies cannot answer that. The fat-burning studies measured hours after one dose, and neither measured fat burning long enough to show how long the drop lasts. The weight studies were mostly snapshots that measured no fat burning and cannot show cause. The review’s authors offer possible reasons for moderate drinkers’ lower odds of obesity: some moderate drinkers may have stopped drinking because of obesity-related illnesses, drinkers may eat less after drinking, or moderate drinkers may lead healthier lives; with snapshot data, they call the cause unclear.
What lab studies did measure is the meal. A review of 22 controlled studies found that people ate on average about 82 more calories of food at a meal or snack after an alcoholic drink than after a non-alcoholic one (12 studies); counting the drink itself, total intake was about 256 calories higher (8 of them). Pooled, the rise in food intake at high doses, 30 g or more, was not significant, though the one study in the review that tested several doses found food intake rose only after 32 g, and the authors call for more research on dose. The review does not report what people ate the next day.
Two popular explanations meet these studies: that alcohol turns into fat, and that it hurts muscle by lowering testosterone. Fat: very little alcohol was turned into fat, as above. Muscle: the muscle study points to signalling inside the muscle: the signalling protein mTOR was less activated two hours after the workout when alcohol was taken, and the authors write that the drop in muscle protein building was “likely mediated, at least in part” by that signalling, a mechanism they call “still poorly defined”. Testosterone was not measured in that study.
A separate review of 17 observational studies of more than 18,000 men found no significant overall difference in testosterone between drinkers and non-drinkers. By intake, men with low and moderate intake had higher testosterone on average, the heavy-intake group showed no significant difference, and a group whose intake was not reported had lower levels. The authors still conclude that alcohol has a “negative significant effect” on testosterone, so this review does not settle the question.
The Dose These Studies Did Not Test
The muscle question is the one these studies answer least.
Ask any fitness forum what two or three beers after the gym does to your gains and you will get confident responses in both directions.
But none of the studies behind this page measured what a moderate dose of alcohol does to muscle protein synthesis. The 24 to 37 percent reduction comes from eight young men who drank about 12 standard drinks, an amount based on binge drinking by athletes. That number tells you what one binge-sized dose did. It tells you nothing about smaller ones.
One training study looked at lean mass instead, with limits. In a 10-week trial, 72 young adults did interval training twice a week, or no training; those who chose alcohol drank beer, or the same alcohol as vodka in sparkling water, with meals on weekdays (one 330 mL beer a day for women, two for men) and at weekends. Lean mass rose about 5% in every training group, drinkers included. The participants chose whether to drink, and the authors say the trial may have been too small to show differences between the drinks. It measured lean mass, not muscle protein synthesis.
Population-level data shows no association between moderate drinking and muscle mass. The one measured dose, and what protein did there, is covered in the evidence on alcohol and muscle protein synthesis.
But the question most people want answered sits in a gap these studies have not closed. That honest gap is more useful than any confident answer built on a twelve-drink experiment.
The Threshold You Are Probably Ignoring
Sleep is the only outcome here where a dose model found where a change begins.
Across the pooled studies, alcohol delayed REM sleep by 18 minutes on average and cut it by 11 minutes. In the dose models, the delay became significant from about 0.35 g per kg and the loss of REM sleep from about 0.50 g per kg, which the paper calls about two standard drinks in its text and about three in its figure.
The review’s authors note that disrupted REM sleep is suggested to impair memory consolidation, cognitive function and emotional regulation. It did not measure appetite, muscle recovery or training the next day.
The full results for every stage of sleep are in the analysis of the 27 sleep studies.
How people felt was not pooled: of five studies that asked about sleep quality, two found it rated worse after alcohol, one better and two no different.
Falling asleep got significantly faster only from about five standard drinks, a dose about 2.4 times the one at which the REM delay became significant; overall, alcohol did not shorten the time to fall asleep significantly. The authors conclude that any benefit of alcohol as a sleep aid is likely outweighed by the disruption to REM sleep.
Sleep gives the clearest number on this page because it is the only outcome with a modelled starting dose, though the authors note that their dose thresholds rest on statistical significance, on the assumption that it marks a clinically significant change. None of these studies linked the sleep changes to muscle or fat.
Your Threshold Depends on Your Priority
So the honest answer depends on which outcome matters most to you, and on what each study could show.
If sleep is your priority, the line the sleep review’s dose model found is about two standard drinks of 10 g in an evening, where the delay to REM sleep became significant. None of these studies measured what that means for muscle or training.
If your main concern is weight, moderate drinking, 14 to 28 g a day, was not linked with higher odds of overweight in the snapshot studies; heavy drinking was. If you log your drinks, the extra food is the part left to watch: about 82 calories at a lab meal or snack, on average.
How a week of drinking compares, seven nights or two, is a question none of these studies tested.
If you stopped drinking because you heard it lowers testosterone, the review here found no significant overall difference, and higher levels on average at low and moderate intake, though its authors conclude the opposite. The one measured cost to muscle protein building came at about 12 drinks, and its authors link it, at least in part, to signalling inside the muscle.
These studies do not give a single number. What they give is what was measured, at what dose, for each outcome.
Whether those costs add up to something incompatible with your fitness goals is the question the companion analysis takes on.
The full guide to alcohol and body composition covers each of these studies.
The answer depends on which outcome you prioritize, and on what was measured.
If you care most about sleep, the sleep review found REM sleep delayed from about two standard drinks of 10 g in an evening. Most of its studies tested a single night, so it does not show what weeks of nightly drinks add up to.
If you care most about your weight, moderate drinking (14 to 28 g a day) was not linked with higher odds of overweight in mostly snapshot studies; heavy drinking was. If you log your drinks, the part left to watch is the food: about 82 extra calories at a lab meal or snack, on average.
If you care most about muscle, none of these studies measured what 2–3 drinks do to protein synthesis; the only data point is at about 12 drinks. One training trial with beer or vodka at meals found lean mass gains similar in drinkers and non-drinkers, but it measured lean mass, not protein synthesis.