A multi-state life-table model of a cohort followed from age 12, parameterized with New Zealand data and using WHO and Global Burden of Disease relative-risk libraries, estimated years of life lost across 0-100 g of daily ethanol. Using WHO risk functions, a 1-in-100 lifetime risk of alcohol-attributable death arrived at 8.6 standard drinks weekly for males and 9.2 for females. Excluding conditions with weak (one-star) evidence cut those thresholds to roughly 2.7 and 3.5 drinks. Māori crossed risk thresholds at lower intake than non-Māori.
The headline lesson is how sensitive the answer is to modelling choices. A threefold swing in the 'safe' limit, depending on which diseases count as causally linked to alcohol, shows that guideline numbers are partly a judgement about evidence quality, not a fixed biological fact. Still, even the most permissive scenario sits near 9 drinks a week, which is below many national guidelines, including older New Zealand advice. That fits the direction of recent guidance from Canada, which now frames risk as low at two drinks weekly or fewer, and from the WHO position that no level is risk-free. Limits are real: this is a simulation, not observed outcomes. It assumes lifelong constant intake, relies on observational relative risks prone to confounding and self-report error, and treats a 17.5-year-loss proxy as a death. Confirmatory rather than paradigm-shifting, it still gives a practical takeaway: lower is safer, and the first few drinks carry measurable cost.