Nassim Taleb critiques a US government-funded study claiming no health benefits from low alcohol consumption. The study, *Alcohol Intake and Health Study*, argues mortality rises to 4% at 14 drinks/week, advising ≤1 drink/day. Taleb identifies statistical flaws: asymmetric confidence intervals (CIs) misrepresent risks (e.g., 7 drinks’ CI should center at -7.01, not -5.47), questionable Monte Carlo methods, and non-transparent correlation assumptions. Table 1’s data shows gains at low consumption, contradicting conclusions. The study’s reliance on a high-variance age group (>70 years) with poor cause-of-death attribution undermines reliability. No sensitivity analysis or bootstrapping was performed. Taleb concludes the study’s methodology is statistically strange and insufficient to overturn decades of alcohol guidelines.

Key critiques focus on the 7 drinks/week and 14 drinks/week mortality thresholds. The 1 in 1,000 and 1 in 25 risk figures are central to the paper’s argument but lack rigorous validation. Taleb highlights the 14 drinks/week group’s dominance in results, where data is sparse and prone to survivor bias. The paper’s lack of transparency in CI calculations and unexplained statistical techniques further weaken its credibility.

Taleb emphasizes the study’s failure to address fundamental statistical principles. The unweighted sum of age-group risks produces inconsistent CIs, suggesting flawed uncertainty propagation. The supplemental materials offer no clarification, leaving methodology opaque. For a study aiming to redefine alcohol recommendations, these flaws render conclusions unreliable. Taleb’s analysis underscores the need for stricter scrutiny of government-sponsored health research.