Reviews/Alcohol and Health/No safe level of alcohol
Liver Mortality · ALC-LM-02 · v2
Borderline: within 4 points of the Uncertain/Doubtful boundary, so a re-grade could land either side.
The claim is an absolute, universal one: that no amount of alcohol is safe. Once the panel set aside the large global-burden syntheses and reviews and kept only direct primary studies, what remained pointed in different directions depending on the outcome. The strongest study [10] found that for stroke and several heart-related conditions there was no level low enough to stop reducing risk, yet it placed the lowest death risk at about 100 grams of alcohol per week rather than zero, and even found alcohol linked to slightly less heart-attack risk. Other surviving studies found harm with no safe level for a blood-chemistry marker [19] and for drinking during pregnancy [21][22], and a regional study linked more drinking to more alcohol-related deaths [12]. Because one direct counter-signal sits inside the best mortality evidence, and the pregnancy findings apply to a specific group and mostly at higher doses, the panel concluded the sweeping universal wording is not carried by what remained.
The surviving primary evidence is outcome-specific. Wood [10], pooling individual data on 599,912 current drinkers across 83 prospective studies, found alcohol roughly linearly associated with higher stroke risk (HR per 100 g per week 1.14, 95% CI 1.10-1.17) and no clear threshold for several cardiovascular subtypes, yet placed the all-cause mortality minimum around or below 100 g per week and found alcohol inversely associated with myocardial infarction (HR 0.94, 0.91-0.97). Ueno [19] reported that even low or occasional intake elevated hyperuricemia risk (uric acid >7.0 mg/dL), concluding no safe level for that endpoint. Prenatal studies [21][22] reported harm, and Robinson [12] found alcohol-related mortality generally higher in regions with higher per-adult consumption. The contested point: The division is whether outcome-specific no-threshold findings carry an absolute, universal statement. The flagship primary study [10] simultaneously supports sub-guideline limits for some outcomes and reports a non-zero mortality minimum plus a protective myocardial-infarction association, which directly undercuts a universal reading.
All 22 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from Alcohol and Health, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
The one direct primary test, Wood [10], contradicts the absolute wording: it places the all-cause mortality minimum 'around or below 100 g per week' (non-zero) and finds alcohol inversely associated with myocardial infarction (HR 0.94), so a universal 'no safe level' is not supported for every outcome.
Papers [21] and [22] are prenatal-exposure studies; 'no safe level during pregnancy' is a narrower population claim and cannot be generalised to all drinkers without conflation.
Where it remains uncertain. The same flagship study [10] places the all-cause mortality minimum around or below 100 g per week and finds alcohol inversely associated with myocardial infarction (HR 0.94), while the no-safe-level support in [21][22] is pregnancy-specific.
Aleatory
Within [10] the dose-response runs opposite ways across outcomes (higher stroke risk but lower myocardial-infarction risk), so a single safe-level verdict cannot hold across all endpoints at once.
Epistemic
The syntheses covering cancer [11] and injury [14] were removed by the study-type filter, so the surviving primary corpus does not speak to large parts of the harm claim.
Inferential
Ueno [19] uses serum uric acid >7.0 mg/dL, a biochemical marker; generalising a surrogate endpoint to a universal safe-level statement is an indirectness leap.
The load-bearing papers from a corpus of 22, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Ecological support for a consumption-harm gradient at population level.
“Alcohol-related mortality was generally higher in regions with higher per adult consumption.”
Panel auditInternal validity: low · scope of inference: Regional population consumption and alcohol-related mortality in Great Britain, 2010-11.Key limitations: ecological fallacy risk · no individual dose-response · cross-sectional
Supports no safe level for the hyperuricemia endpoint (a biochemical surrogate).
“Even low or occasional intake elevated the risk, indicating no safe level of alcohol consumption.”
Panel auditInternal validity: medium · scope of inference: Risk of incident hyperuricemia in Japanese men and women; not all-cause health.Key limitations: single biochemical outcome · self-reported consumption · one national population
Shows prenatal harm but at moderate-to-high, not low, exposure.
“infants with moderate-to-high PAE had 0.17 SD (95% CI -0.30, -0.04) lower cognitive and 0.15 SD (-0.29, -0.2) lower expressive language scores compared to infants without PAE”
Panel auditInternal validity: medium · scope of inference: Effect of moderate-to-high prenatal alcohol exposure on 1-year child development in two settings.Key limitations: no low-level exposure contrast · confounded by socioeconomic factors · single early time point
Primary support: for most CVD subtypes no lower safe threshold is observed.
“increased alcohol consumption was log-linearly associated with a lower risk of myocardial infarction (HR 0·94, 0·91-0·97)”
Panel auditInternal validity: high · scope of inference: Current drinkers without prior cardiovascular disease in 19 high-income countries; does not include never-drinkers as a comparator.Key limitations: current drinkers only, no abstainer reference · observational confounding · mixed direction across outcomes
4 of 22 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
Outcome-specific primary evidence leans toward harm with thresholds below guidelines, but a direct counter-signal in the flagship mortality analysis and the pregnancy-specific, dose-limited nature of the supporting studies prevent an absolute universal verdict.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Low-dose harm in the general population. The claim requires harm at low intake, but [10] shows a non-zero mortality minimum and the low-dose no-safe-level signals [19][22] are a surrogate and pregnancy-specific. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: A large prospective general-population study with hard clinical endpoints resolving the dose-response below 100 g per week, including genuine abstainers as the reference.
Open question: Does all-cause mortality risk rise from zero intake, or is there a non-zero minimum in the general adult population? Why it is still open: The one direct test [10] places the minimum around or below 100 g per week, not at zero, and finds alcohol inversely associated with myocardial infarction (HR 0.94).
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 21 Oct 2026 · protocol v0.19