Alcohol before bed delays and shortens REM sleep from about two standard drinks, and the effect grows with dose; total sleep time and sleep efficiency are not reliably changed
Key Findings
REM sleep onset was delayed and REM duration reduced already at a low dose (≤0.50 g/kg, about two standard drinks), worsening progressively with higher doses.
27 studies of alcohol taken before bedtime in healthy adults. Low dose was defined as ≤0.50 g/kg (about two standard drinks), high dose as ≥0.85 g/kg (about five).
REM sleep: onset delayed and duration reduced. The disruption was already present at the low dose and worsened progressively with higher doses.
Shorter sleep onset latency and shorter latency to deep sleep (N3) appeared only at the high dose.
Total sleep time, sleep efficiency and wake after sleep onset: large uncertainty, no reliable effect.
Practical Application
Two drinks in the evening already cost REM sleep; five cost more. Falling asleep faster after drinking is not a sign of a better night — it appears only at high doses and comes with the largest REM loss.
Do not tell an athlete that a drink shortens their sleep or lowers sleep efficiency: this meta-analysis could not show it.
Before a key session or race the lever is the dose, and the evidence has no safe threshold above zero for REM; frame it as a trade the athlete chooses, not a prohibition.
Limitations
Healthy adults, not athletes; effect sizes are not given in the abstract (abstract only was read).
Sex differences were flagged by the authors as unstudied.
An older narrative review (Ebrahim 2013, DOI 10.1111/acer.12006) reported faster sleep onset at all doses; this pooled analysis finds it only at ≥0.85 g/kg and is preferred.