Alcohol consumption and mortality after mitigating sick‐quitter bias: A 19‐year nationwide cohort study of Korean adults

BACKGROUND AND AIM: The lower mortality risk often observed among low-level drinkers remains controversial because estimates may be distorted by sick-quitter bias and poor baseline health among abstainers. We examined whether a dose-response association between alcohol consumption and mortality persisted among drinking groups after applying several strategies to reduce these biases in a nationwide cohort of Korean adults. DESIGN: Prospective nationwide cohort study with sensitivity analyses across four analytic cohorts. SETTING: South Korea. PARTICIPANTS: Using the Korea National Health Examination Baseline cohort, we identified 6 048 052 adults who underwent national health screening in 2004-2005. After excluding participants with major disease history, discordant self-reported chronic disease history between the 2002-2003 and 2004-2005 screening questionnaires, reduced alcohol intake compared with a prior screening or a new cancer diagnosis within the first year of follow-up, 1 103 981 individuals were included in the primary analytic cohort and followed until 31 December 2023 (median follow-up: 19.2 years). MEASUREMENTS: Alcohol intake was categorized as current abstainer, occasional, light, moderate, heavy or very heavy drinking. Because lifetime abstainers could not be distinguished from former drinkers, occasional drinkers were used as the reference group. Outcomes were all-cause, cancer and alcohol-related disease mortality. RESULTS: During follow-up, 50 429 all-cause deaths, 20 110 cancer deaths and 1217 alcohol-related disease deaths occurred. Among drinking groups, mortality risk increased with higher alcohol consumption across all outcomes. For all-cause mortality, adjusted hazard ratios (aHRs) (95% confidence intervals) compared with occasional drinkers were 1.146 (1.108-1.185) for light drinkers, 1.330 (1.281-1.381) for moderate drinkers and 1.954 (1.825-2.092) for very heavy drinkers. For all-cancer mortality, corresponding aHRs were 1.098 (1.043-1.156), 1.230 (1.160-1.304) and 1.714 (1.533-1.917) for light, moderate and very heavy drinkers, respectively. Alcohol-related disease mortality showed the steepest dose-response gradient, with aHRs of 1.962 (1.464-2.628), 3.893 (2.889-5.247) and 14.200 (10.020-20.124) for light, moderate and very heavy drinkers. Furthermore, abstainers had lower risk than occasional drinkers for alcohol-related disease mortality [0.525 (0.361-0.762)]. Sensitivity analyses showed that estimates for abstainers varied materially across outcomes and analytic cohorts, with the most pronounced instability observed for alcohol-related disease mortality. SUMMARY: Mortality risk appears to increase in a consistent dose-dependent fashion across drinking groups in Korean adults, with even light drinking associated with higher risk than occasional drinking across all major outcomes. These findings do not support a protective interpretation of low-level alcohol consumption and suggest that alcohol-related risk begins at low levels of intake, particularly for cancer and alcohol-related outcomes.