Alcohol

The "moderate drinking is healthy" story has not survived rigorous methods. Mendelian randomization shows monotonic harm at every dose; cancer risk begins at the first drink. The minimum-risk amount for a healthy adult is at or near zero — and people who don't drink should not start.

The scientific consensus on alcohol has shifted substantially over the past decade. The historical "J-curve" — a few drinks per day appearing protective for cardiovascular events and all-cause mortality — has largely collapsed under more rigorous methods, particularly Mendelian randomization and meta-analyses correcting for "abstainer bias."

The current evidence-based view: the minimum-risk dose for a healthy adult is at or near zero. People who don't drink should not start.

The shift: 2023–2025 official statements

WHO January 2023: "No level of alcohol consumption is safe for our health."^[1]
Canadian Centre on Substance Use 2023: revised guidance to "less is better"; <2 drinks/week as lower-risk threshold.
U.S. Surgeon General January 2025 Advisory on alcohol and cancer: alcohol identified as third leading preventable cause of cancer in the US (after tobacco and obesity); calls for updated cancer warnings.^[2]
NASEM 2025 (National Academies of Sciences) Review remains less hawkish — found moderate drinkers had lower all-cause and CVD mortality vs. abstainers in studies excluding former drinkers — illustrating the field is not fully unanimous.

The trend is unambiguous: harm signals dominate, the J-curve is most likely an artifact of "sick quitter" bias, and recommended limits keep falling.

Cancer: the strongest harm signal

IARC classifies alcoholic beverages as Group 1 carcinogen (since 1988); acetaldehyde from alcohol added as Group 1 in 2009 — same category as tobacco and asbestos.

Sufficient evidence that alcohol causes seven cancer types:

Oral cavity
Pharynx
Larynx
Esophagus (squamous cell)
Liver
Colorectum
Female breast

Approximately 4% of cancers worldwide are attributable to alcohol (~740,000 cases/year). In the US: ~100,000 cases and 20,000 deaths annually (2025 Surgeon General Advisory).

Dose-response

A landmark dose-response meta-analysis (572 studies, 486,538 cases) showed clear monotonic dose-response curves.^[3] Even at light drinking (≤1 drink/day, ~12.5 g ethanol):

Oral/pharyngeal cancer: RR 1.13
Esophageal SCC: RR 1.26
Female breast cancer: RR 1.04

At moderate drinking (~25 g/day):

Oral/pharyngeal: RR 1.83
Esophageal SCC: RR 2.23
Breast: RR 1.23
Colorectal: RR 1.21

Breast cancer specifically: dose-response increases starting from <5 g/day; 10.5% increase per 10 g/day. For a woman with average breast cancer risk, 1 drink/day raises lifetime risk meaningfully.

Mechanisms

Acetaldehyde — first oxidative metabolite of ethanol; forms DNA adducts; itself Group 1 carcinogen
Reactive oxygen species from CYP2E1-mediated metabolism — oxidative DNA damage
Hormonal effects — alcohol elevates circulating estrogens (mechanism for breast cancer)
Disrupted folate / one-carbon metabolism — impairs DNA methylation
Impaired absorption of protective nutrients (vitamins A, D, E, folate, carotenoids)
ALDH2*2 East Asian variant (impaired acetaldehyde clearance) — markedly elevated UADT cancer risk in carriers who drink: genetic confirmation of causality

Cardiovascular: the J-curve has collapsed

For three decades, observational studies showed a J-shape: light drinkers (~1 drink/day) appeared healthier than abstainers. The cardio-protective hypothesis was anchored on alcohol's HDL-raising and modest anti-platelet effects.

Mendelian randomization deflates the J-curve

Using genetic variants (most importantly ADH1B rs1229984 and ALDH2 rs671) as unconfounded proxies for lifetime alcohol exposure:

A China Kadoorie Biobank study (n=512,000 Chinese adults) found any genetically predicted increase in alcohol intake raised BP and stroke incidence linearly, with no protective threshold.^[4]31772-0/fulltext)
A 94-SNP MR analysis: OR 1.27 for stroke, OR 3.05 for peripheral artery disease per 1-SD increase in log-drinks/week.^[5]
A UK Biobank MR: each additional daily drink causally associated with +2.65 mmHg systolic BP, OR 2.25 for hemorrhagic stroke, OR 1.26 for atrial fibrillation.^[6]
A 2024 Million Veteran Program analysis replicated the U-shaped observational curve, but in MR found no causal benefit at moderate intake.^[7]

Specific cardiovascular outcomes

Hypertension: linear positive dose-response, no threshold.^[8]
Atrial fibrillation: ~10% increase per drink/day (Tu et al. 2024).
Stroke: increases with intake.
Cardiomyopathy: heavy chronic drinking causes dilated cardiomyopathy.

The remaining cardio-protective signal in some studies appears to be confounded by socioeconomic and lifestyle factors that travel with moderate drinking in some populations.

Liver disease

Direct dose-response for alcoholic fatty liver disease, alcoholic hepatitis, and cirrhosis. No "safe" threshold. Heavy chronic drinking is the dominant cause of cirrhosis worldwide alongside hepatitis viruses. Even moderate drinking accelerates the progression of MASLD (formerly NAFLD).

Brain

Alcohol use disorder clearly causes brain atrophy and cognitive decline.
Even moderate drinking is associated with reduced gray matter and hippocampal volume in cohort studies.^[9]^[10]
2024 Lancet Commission on Dementia includes excessive alcohol as a modifiable risk factor.
The "moderate alcohol protects against dementia" claim from older studies has not survived rigorous methods.

Mental health

Alcohol is a depressant; chronic drinking exacerbates depression and anxiety.
Acute use can suppress sleep architecture (REM suppression, fragmented sleep).
Heavy drinking is a major risk factor for suicide.
Even moderate use degrades sleep quality.

Sleep specifically

Alcohol hastens sleep onset but fragments architecture:
- Suppresses REM in first half of night
- Causes wakefulness in second half (rebound)
- Worsens sleep apnea
Even 1–2 drinks within 3 hours of bed measurably reduces sleep quality.

Practical: if you drink, do so well before bed (≥3 hours), or avoid on poor-sleep nights.

Pregnancy

No safe level. FAS spectrum disorders are dose-dependent and individual susceptibility varies widely. Recommendation: complete avoidance during pregnancy and ideally pre-conception.

Quantifying "a drink"

Standard drink definitions vary by country. US standard drink = 14 g pure ethanol = ~12 oz beer (5%) = 5 oz wine (12%) = 1.5 oz spirits (40%). UK unit = 8 g ethanol. Czech standard varies.

Modern guidance:

Lowest risk: 0 drinks/week
Lower risk (Canada 2023): <2 drinks/week
Increasingly risky: 3–6 drinks/week
High risk: 7+ drinks/week (or any single-occasion >4 in women, >5 in men)

The shift from "moderate is fine" to "less is better, ideally zero" reflects evolving evidence, not new puritanism.

Practical guidance

If you don't drink, don't start for "health benefits" — there are none that survive rigorous methods.
If you drink, less is better. Treat each drink as an exposure.
Avoid binge drinking (≥4 drinks for women, ≥5 for men in 2 hours) — particularly harmful for liver, brain, and cardiovascular acute events.
Don't drink within 3 hours of bed if sleep matters.
Family history of breast cancer: the breast cancer dose-response starts at <5 g/day. Consider this seriously.
East Asian heritage with flushing reaction (ALDH2*2 carriers): drinking carries dramatically elevated UADT cancer risk. Avoidance is appropriate.
Pregnancy, planning pregnancy, breastfeeding: abstain.
Use periods of abstinence to assess your own dependency. Difficulty stopping is a flag.
Cancer screening doesn't fully compensate. Reducing alcohol is more effective than catching cancer earlier.

What about red wine and resveratrol?

A 5-oz glass of red wine contains 0.3–1 mg resveratrol.
Resveratrol research uses doses of 100–500 mg.
The cardiovascular signal in old "wine drinkers live longer" cohorts is overwhelmingly explained by lifestyle confounding — wine drinkers in those studies tended to be wealthier, more educated, more active, with higher diet quality.
There is no special protective effect of wine over other alcohol that survives rigorous methods.