| Popis: |
Background Observational epidemiological studies have consistently demonstrated a lower risk of ischaemic heart disease (IHD) and ischaemic stroke (IS) among moderate compared with non-drinkers, and heterogeneous associations of heavy alcohol intake with these cardiovascular disease (CVD) outcomes. However, the causal nature of these associations remains controversial, and potential mechanisms are not well understood. Methods The UK Biobank is a prospective study of 502,536 adults aged 40-69 years enrolled between 2006 and 2010. For observational analyses, 180,565 White male participants free of baseline CVD were included in Cox regression analyses to estimate risk of incident IHD or IS associated with self-reported baseline alcohol intake, to explore the impact of measurement error, confounding, and reverse causality bias, accounting for these where possible, and to estimate mediating pathways. A Mendelian randomisation (MR) analysis was conducted among 148,488 unrelated male participants with White British ancestry using an externally-weighted genetic instrument based on 93 SNPs associated with alcohol intake. Logistic regression was used to estimate the associations of genetically-predicted log alcohol intake with incident and prevalent IHD and IS. Localised average causal effect (LACE) estimation was used to assess linearity of the MR associations. Results The mean (SD) age of participants was 57 (8) years, and 80% consumed alcohol at least weekly. The age-adjusted hazard ratio (HR) for the association of lifetime abstention versus moderate alcohol intake (0.1-14 units/week) with IHD (1918 events) was attenuated from 1.27 (95%CI 1.06-1.51) to 1.11 (0.85-1.45) after adjustment for confounding by SES, smoking, diet, physical activity, and region, and exclusion of sources of reverse causality bias including reduced-intake drinkers and those with long-standing illness or disability; for IS (499 events), the corresponding HR increased from 1.31 (0.94-1.83) to 1.74 (1.10-2.73), but there were too few events for reliable conclusions. The association of heaviest (>42 units/week) versus moderate alcohol intake with IHD was null (0.92; 0.80-1.07), and with IS was positive (1.38; 1.05-1.82); these associations were mediated mainly by an inverse mediating effect of HDL-c (stronger for IHD than IS) and positive mediating effects of systolic blood pressure (SBP), body mass index (BMI) and, for IS, atrial fibrillation (AF). The MR analyses provided weak evidence that higher alcohol intake causally increases risk of IHD (OR per doubling of genetically-predicted alcohol intake 1.09, 95% CI 1.01-1.19, p=0.035), with borderline evidence of non-linearity (p=0.027) (14,253 events). There was no evidence of a casual association between alcohol intake and risk of IS (OR per doubling of genetically-predicted alcohol intake 1.08, 95% CI 0.90-1.30, p=0.383) but this analysis was not well-powered (2,651 events). Conclusion Classical observational and MR analyses support the hypothesis that the lower observed risk of CVD among moderate drinkers compared with non-drinkers is not causal, and that the association is largely attributable to residual observational biases, particularly reverse causality bias related to longstanding illness or disability among lifetime abstainers. Heterogeneous associations of heavy alcohol intake with IHD and IS may be attributable to differential mediating effects, including those of HDL-c and AF (although the possibility of a causal role for HDL-c is uncertain). |
