What the Research Says About Alcohol

One glass of wine with dinner improved sleep quality in people with diabetes — but the same dose measurably slowed your brain's processing speed

Here’s the tension at the heart of the alcohol research: a 150 mL glass of red wine with dinner, consumed daily for two years, improved self-reported sleep quality in adults with type 2 diabetes, according to a randomized controlled trial. But a meta-analysis of 62 studies found that the same dose range (roughly 0.3 g/kg, or about one drink for a 70 kg person) reduced the amplitude of the P300 brain wave by 1–2 microvolts — a medium-to-large effect (Hedges’ g ≈ 0.6–0.8) — indicating slower, less efficient neural processing of unexpected stimuli. That’s the problem: the same substance that helps you fall asleep is also, at the same dose, dulling your brain’s ability to react to the world. Most wellness advice ignores this trade-off entirely.

What the research actually shows

The most credible evidence on moderate alcohol comes from randomized controlled trials, not observational studies. The CASCADE trial (n=224) is the standout: alcohol-abstaining adults with well-controlled type 2 diabetes were randomized to 150 mL of red wine, white wine, or mineral water with dinner daily for two years, all while following a Mediterranean diet. The red wine group showed modest improvements in lipid profiles (higher HDL cholesterol) and self-reported sleep quality compared to the mineral water group. Importantly, there were no significant differences in blood pressure, liver enzymes, or medication use between groups — meaning the intervention was safe in this specific population over two years.

But the neural data tells a different story. The meta-analysis of 62 studies (n=1,948 participants) found that acute alcohol consumption consistently reduced P300 amplitude by 1–2 microvolts. This effect was dose-dependent: higher doses (0.6–1.2 g/kg, or 2–5 drinks) produced larger reductions. The N100 component, reflecting early sensory processing, was also reduced. These effects occurred at blood alcohol concentrations well below the legal driving limit in most countries — meaning you don’t need to feel drunk to have measurably impaired neural processing.

The observational data on long-term health outcomes is more mixed and less reliable. The Danish 1905-Cohort Survey (n=2,249) found that in people aged 92–93, alcohol consumption no longer predicted 15-month mortality — traditional risk factors simply stopped mattering at extreme old age. The CDC opioid guidelines and AHA coronary disease guidelines both recommend alcohol moderation, but neither cites strong RCT evidence for cardiovascular benefit. The AASLD liver disease guidance is more direct: drinking more than 40 g/day for women or 60 g/day for men significantly increases cirrhosis risk, and severe alcoholic hepatitis carries a 28-day mortality of 20–50%.

The nuance most people miss

The biggest confound in the alcohol literature is that most observational studies compare “moderate drinkers” to “non-drinkers” — but non-drinkers include former heavy drinkers who quit for health reasons, people too sick to drink, and lifetime abstainers. This creates a systematic bias that makes moderate drinking look protective when it may not be. The CASCADE trial avoids this by only including people who were abstinent at baseline, but it’s one trial in a specific population (diabetics on a Mediterranean diet).

The second nuance: alcohol’s effects depend heavily on context. The CASCADE trial had participants drink with dinner, not on an empty stomach. Food slows alcohol absorption, reducing peak blood alcohol concentration. The neural meta-analysis used fasting or near-fasting conditions in most studies, which amplifies the effect. If you drink with a meal, the P300 reduction is likely smaller — but no study has directly tested this.

The third nuance: individual genetics matter. The CASCADE trial genotyped participants for ADH1B alleles, which affect alcohol metabolism rate. Fast metabolizers showed different lipid responses than slow metabolizers. The neural meta-analysis didn’t account for genetics, so the average effect may not apply to you if you’re a fast or slow metabolizer.

Finally, the liver guidance makes clear that risk thresholds are not symmetrical. The dose that provides a small sleep benefit (one drink) is far below the dose that causes measurable liver damage (40+ g/day for women). But the neural effects occur at the same dose as the sleep benefit — so you’re trading cognitive processing speed for sleep quality, not getting a free lunch.

Practical implications

If you drink, do it with food, not on an empty stomach. The CASCADE trial had participants drink 150 mL of wine with dinner. Food slows alcohol absorption, reducing peak blood alcohol concentration and likely blunting the P300 reduction. A glass of wine on an empty stomach will produce a larger neural effect than the same glass with a meal.

Track your sleep quality and your cognitive performance separately. The CASCADE trial found improved self-reported sleep quality with red wine, but the neural meta-analysis found impaired processing speed at the same dose. If you’re testing alcohol, measure both — a subjective sleep improvement might come with an objective cognitive cost you don’t notice until you measure it.

Stay below 40 g/day for women and 60 g/day for men to avoid liver risk. The AASLD liver guidance is clear: above these thresholds, cirrhosis risk increases significantly. One standard drink contains about 14 g of alcohol, so that’s roughly 2–3 drinks for women and 4–5 for men. The CASCADE trial used 150 mL of wine, which is about 15–18 g of alcohol — well within the safe range.

Don’t start drinking for health benefits if you don’t already drink. The CASCADE trial showed modest improvements in a specific population (diabetics on a Mediterranean diet), but the neural effects are real and immediate. The AHA coronary disease guideline does not recommend starting alcohol for cardiovascular benefit. The net risk-benefit calculation is not clearly positive for non-drinkers.

Design your own experiment

What to test: The effect of one standard drink (150 mL wine, 355 mL beer, or 44 mL spirits) consumed with dinner on next-morning cognitive performance and subjective sleep quality.

How long to run it: Minimum 3 weeks. Run 1 week of baseline measurement (no alcohol), then 2 weeks of the intervention (one drink with dinner). The CASCADE trial ran for two years, but you’ll see measurable effects on sleep within days and cognitive effects within hours. Two weeks gives you enough data points to average out day-to-day variability.

What to measure: Two primary metrics. (1) Next-morning cognitive performance: use a validated online test like the Psychomotor Vigilance Task (PVT) or a simple reaction time test. Measure your average reaction time and lapse frequency within 30 minutes of waking. (2) Subjective sleep quality: use the Pittsburgh Sleep Quality Index (PSQI) or a simple 1–10 scale each morning. Secondary metric: how long it took you to fall asleep (sleep latency).

What confound to watch for: Meal timing and composition. The CASCADE trial had participants drink with dinner, not after. If you drink on an empty stomach or with a high-sugar meal, you’ll get different results. Also watch for caffeine: if you have coffee with dinner or within 4 hours of bed, it will confound both sleep and cognitive measures. Track your caffeine timing separately.

What a positive result looks like: A consistent improvement in subjective sleep quality (at least 1 point on a 10-point scale) with no more than a 5% increase in next-morning reaction time or lapse frequency. If your reaction time increases by more than 5%, the sleep benefit is coming at a cognitive cost you can measure. If your reaction time doesn’t change, you may be a fast metabolizer or the food effect is blunting the neural impact. If your sleep quality doesn’t improve, the trade-off isn’t worth it — and you have your answer.