The Honest Bottom Line

Strip the alcohol debate to its two loudest claims and you get an apparent contradiction. One camp says there is no safe level — risk rises from the first drink. The other shows mortality curves that stay nearly flat until roughly 100 g of ethanol a week. Both are reading real data correctly. This page reconciles them, endpoint by endpoint, and hands you the decision the numbers actually support.

Two Claims, Both Correctly Read

The first claim comes from the Global Burden of Disease analysis: in 2016, no level of drinking improved health in aggregate — the risk-minimizing dose was zero, any protective hint for ischemic heart disease offset by rising cancer risk at every intake (The Lancet, 2018). The World Health Organization puts the same conclusion flatly: no level of alcohol consumption is safe for our health (WHO, 2023). The second comes from 599,912 current drinkers in 83 prospective studies: for all-cause mortality, risk sat on a nearly flat shelf up to about 100 g of ethanol a week — roughly seven US standard drinks — before bending upward (The Lancet, 2018). The reconciliation: the two statements answer different questions. No-safe-level describes direction — risk trends up from zero across the scored endpoints. The shelf describes magnitude for one endpoint, total mortality, where low-dose increments are small enough that the curve looks level. Pull direction from size and the public argument mostly collapses. How the older J-curves got cleaned up is the J-curve saga's story.

Where the Zero-Drink Verdict Comes From

The Global Burden of Disease team linked consumption data from 195 countries to 23 alcohol-related outcomes, from injuries and tuberculosis to cirrhosis, cardiovascular disease, and cancers. For adults over a year, moving from not drinking to one standard drink a day — 10 g of ethanol, smaller than the 14 g US standard — raised the count developing one of those problems from 914 to 918 per 100,000 (The Lancet, 2018). Four extra cases per 100,000 per year: the entire low-dose signal — tiny in size, consistent in direction. What pushes the risk-minimizing dose to zero is the endpoint mix.

• 🎗️ Cancer does not negotiate: the cancer associations rise with dose across most sites, with no threshold identified across 235 studies and more than 117,000 cases (British Journal of Cancer, 2001).
• ❤️ The offset is narrow: the one favorable signal, myocardial infarction (hazard ratio 0.94 per 100 g a week), appears in cohort arithmetic that Mendelian randomization has not confirmed — the genetics point flat-to-up, not down (The Lancet, 2018; BMJ, 2014).
• 🌍 The verdict is population-scale: the GBD optimizes total health loss across everyone, so a small per-person bump across billions of drinkers settles the global question at zero.

Cancer: The Linear Edge of the Curve

Cancer is where the dose-response looks most cleanly linear, and the reason is mechanistic: ethanol metabolizes to acetaldehyde — which IARC lists as carcinogenic to humans in the context of drinking, and which damages DNA directly — so each drink delivers another exposure. A process with no biological off-ramp produces no threshold, and the epidemiology cooperates. A meta-analysis of 222 articles found light drinking — up to one drink a day — already carried elevated risk of oropharyngeal cancer, esophageal squamous cell carcinoma, and female breast cancer (Annals of Oncology, 2013). For breast cancer, the reanalysis of 53 studies — 58,515 cases — found about 7 percent higher risk per additional 10 g of ethanol a day, near enough to one drink (British Journal of Cancer, 2002). Two US cohorts tracking over 135,000 adults found light-to-moderate intake raised alcohol-related cancer risk in women regardless of smoking; in men the signal appeared mainly among smokers (BMJ, 2015).

Small but Real: Reading Both Numbers

A relative risk of 1.05 is easy to weaponize and easy to dismiss, so hold both numbers at once. Per 100,000 people per year, one daily drink produces four extra alcohol-related cases — a rounding error in one person's year, and a real total across billions of drinkers. At the other end the stakes stop being subtle: versus drinking under 100 g a week, those reporting 100–200 g lost about 6 months of life expectancy at age 40, those at 200–350 g lost 1–2 years, and above 350 g — roughly 3–4 US drinks a day — the loss reached 4–5 years (The Lancet, 2018). A Cambridge statistician's counterpoint still stands: there is no risk-free level of driving either.

What This Means for Your Glass

• 🚫 Do not start for health: once cancer enters the ledger, no endpoint shows a net benefit at any dose, and the heart hint does not survive the genetic checks.
• 🍷 If you drink lightly — under about seven US drinks a week: the mortality curve is nearly flat and the added cancer risk is small but one-directional, so the decision can rest on taste, ritual, and sleep rather than a cliff edge.
• ✂️ If you sit above 100 g a week: you live on the steep part of the curve — blood pressure falls within weeks and liver fat largely resolves in one to two months, as the measured wins from cutting show.
• 👩 Sex changes the weights: for women the same dose raises breast-cancer risk linearly — alcohol's amplified cost prices that asymmetry.

Watch-Items and Honest Limits

• ⚖️ Association is still association: the big curves adjust for smoking, age, and diabetes, but adjustment is not randomization; the genetic studies narrow the uncertainty without closing it.
• 🧮 Population curves price cohorts, not you: a hazard ratio per 100 g describes thousands of drinkers on average — not one person's next decade.
• ⏳ Cancer latency runs in decades: today's cancer curves partly reflect drinking from 20–30 years ago; today's reduction takes years to show.
• 🧠 The brain runs a separate ledger: cognition and structural change price differently from metabolism — alcohol and the brain owns that accounting.

Questions, Answered Briefly

• 🍷 "Is a glass of wine with dinner bad for me?" It carries a small, one-directional cancer increment and no established health benefit. Below seven drinks a week, reasonable people decide differently — but decide knowing both numbers.
• 📉 "Is cancer risk really linear with no threshold?" It is the best-supported reading: the acetaldehyde mechanism implies no floor, and the pooled epidemiology has failed to find one for most alcohol-related cancers (British Journal of Cancer, 2001). Linear does not mean steep — the smallest doses add the smallest increments.
• ⏱️ "Zero drinks or fewer drinks?" Zero is the risk minimum. But most measurable harm lives above 100 g a week, so the steepest gains come from cutting there — and the evening pour can be re-priced for sleep, as the alcohol calculus does.

Related Topics